It's taken a while to get here, but the code of conduct that I set out to produce about a year ago now has been voted on. The result which has been released earlier this week is that the code of conduct has been accepted, but that future updates would require another GR. Although not my personal preference, it's pretty much what I expected. I still think allowing the DPL to update the code of conduct in the future would be preferable over having to go through another GR (after all, if the DPL does something silly, we have ways of reverting that decision); but regardless, the fact that the code of conduct is accepted is, in my opinion, a major step forward for Debian. Thanks to everyone who voted!

I have heard many good things about Docker, and decided to give it a spin on my systems. I think application-level virtualization has a lot to offer to my workflow... But the process to understand and later adopt it has left me somewhat heart-torn. Docker is clearly great technology, but its documentation is... Condescending and completely out of line with what I have grown used to in my years using Linux. First, there is so much simplistic self-praise sprinkled throughout it. There is almost no page I landed on that does not mention how user-friendly and user-centric Docker's commandline arguments are They let you talk in almost plain1 English. What they don't mention is that... Well, that's the way of basically every command-line tool. Of course, as soon as you start specifying details to it, the plain-Englishness starts dilluting into a more realistic English-inspiredness... Then... Things that go against our historical culture. It is often said that Windows documentation tends to be repetitive because users don't have the patience to read a full document. And our man pages are succint and to the point, because in our culture it is expected that users know how to search for the bit of information they are after. But reading documentation that's so excited with itself and praises again and again the same values and virtues, but never gets to the point I am interested in getting at (be it deployment, interoperation, description of the in-disk images+overlays layout, or anything moderately technical) never gets there... makes me quite unhappy. Last (for now)... Such a continuous sales pitch, an insistence on the good virtues, makes me wary of something they might be hiding. Anyway, at least for now, I just wanted to play a bit with it; I will wait at least until there is a backport to the stable Debian version before I consider moving my LXC VMs setup over to Docker (and a backport does not seem trivial to achieve, as Docker has several updated low-level dependencies we are unlikely to see in Wheezy). But I had to vent this. OK, now go back to your regular work ;-)

1. Of course, plain as long as you agree to a formal grammar... Details, details

Background I ve previously written about the claim that people use Autism as an excuse for bad behavior [1]. In summary it doesn t and such claims instead lead to people not being assessed for Autism. I ve also previously written about empathy and Autism in the context of discussions about conference sexual harassment [2]. The main point is that anyone who s going to blame empathy disorders for the widespread mistreatment of women in society and divert the subject from the actions of average men to men in minority groups isn t demonstrating empathy. Discussions of the actions of average men are so often derailed to cover Autism that the Geek Feminism Wiki has a page about the issue of blaming Autism [3]. The Latest Issue Last year Shanley Kane wrote an informative article for Medium titled What Can Men Do about the treatment of women in the IT industry [4]. It s a good article, I recommend reading it. As an aside @shanley s twitter feed is worth reading [5]. In response to Shanley s article Jeff Atwood wrote an article of the same title this year which covered lots of other things [6]. He writes about Autism but doesn t seem to realise that officially Asperger Syndrome is now Autism according to DSM-V (they decided that separate diagnosis of Autism, Asperger Syndrome, and PDD-NOS were too difficult and merged them). Asperger Syndrome is now a term that refers to historic issues (IE research that was published before DSM-V) and slang use. Gender and the Autism Spectrum Jeff claims that autism skews heavily towards males at a 4:1 ratio and cites the Epidemiology of Autism Wikipedia page as a reference. Firstly that page isn t a great reference, I fixed one major error (which was obviously wrong to anyone who knows anything about Autism and also contradicted the cited reference) in the first section while writing this post. The Wikipedia page cites a PDF about the Epidemiology of Autism that claims the 4.3:1 ratio of boys to girls [7]. However that PDF is a summary of other articles and the one which originated the 4.3:1 claim is behind a paywall. One thing that is worth noting in the PDF is that the section containing the 4.3:1 claim also references claims about correlations between race and Autism and studies contradicting such claims it notes the possibility of ascertainment bias . I think that anyone who reads that section should immediately consider the possibility of ascertainment bias in regard to the gender ratio. Most people who are diagnosed with Autism are diagnosed as children. An Autism diagnosis of a child is quite subjective, an important part is an IQ test (where the psychologist interprets the intent of the child in the many cases where answers aren t clear) to compare social skills with IQ. So whether a child is diagnosed is determined by the psychologist s impression of the child s IQ vs the impression of their social skills. Whether a child is even taken for assessment depends on whether they act in a way that s considered to be obviously different. Any child who is suspected of being on the Autism Spectrum will be compared other children who have been diagnosed (IE mostly boys) and this will probably increase the probability that a boy will be assessed. So an Aspie girl might not be assessed because she acts like other Aspie girls not like the Aspie boys her parents and teachers have seen. The way kids act is not solely determined by neuro-type. Our society expects and encourages boys to be louder than girls and take longer and more frequent turns to speak, this is so widespread that I don t think it s possible for parents to avoid it if their kids are exposed to the outside world. Because of this boys who would be diagnosed with Asperger Syndrome by DSM-IV tend to act in ways that are obviously different from other kids. While the combination of Autism and the the social expectations on girls tends to result in girls who are quiet, shy, and apologetic. The fact that girls are less obviously different and that their differences cause fewer difficulties for parents and teachers makes them less likely to be assessed. Note that the differences in behavior of boys and girls who have been diagnosed is noted by the professionals (and was discussed at a conference on AsperGirls that my wife attended) while the idea that this affects assessment rates is my theory. Jeff also cites the book The Essential Difference: Male And Female Brains And The Truth About Autism by Professor Simon Baron-Cohen (who s (in)famous for his Extreme Male Brain theory). The first thing to note about the Extreme Male Brain theory are that it depends almost entirely on the 4.3:1 ratio of males to females on the Autism Spectrum (which is dubious as I noted above). The only other evidence in support of it is subjective studies of children which suffer from the same cultural issues this is why double blind tests should be used whenever possible. The book Delusions of Gender by Cordelia Fine [8] debunks Simon Baron-Cohen s work among other things. The look inside feature of the Amazon page for Delusions of Gender allows you to read about Simon Baron-Cohen s work [9]. Now even if the Extreme Male Brain theory had any merit it would be a really bad idea to cite it (or a book based on it) if you want to make things better for women in the IT industry. Cordelia s book debunks the science and also shows how such claims about supposed essential difference are taken as exclusionary. The Problem with Jeff Atwood Jeff suggests in his post that men should listen to women. Then he and his followers have a huge flame-war with many women over twitter during which which he tweeted Trying to diversify my follows by following any female voices that engaged me in a civil, constructive way recently . If you only listen to women who agree with you then that doesn t really count as listening to women. When you have a stated policy of only listening to women who agree then it seems to be more about limiting what women may feel free to say around you. The Geek Feminism wiki page about the Tone Argument [10] says the following: One way in which the tone argument frequently manifests itself is as a call for civility. A way to gauge whether a request for civility is sincere or not is to ask whether the person asking for civility has more power along whatever axes are contextually relevant (see Intersectionality) than the person being called incivil , less power, or equal power. Often, people who have the privilege of being listened to and taken seriously level accusations of incivility as a silencing tactic, and label as incivil any speech or behavior that questions their privilege. For example, some men label any feminist thought or speech as hostile or impolite; there is no way for anybody to question male power or privilege without being called rude or aggressive. Likewise, some white people label any critical discussion of race, particularly when initiated by people of color, as incivil. Writing about one topic is also a really good idea. A blog post titled What Can Men Do should be about things that men can do. Not about Autism, speculation about supposed inherent differences between men and women which are based on bad research, gender diversity in various occupations, etc. Following up a post on What Can Men Do with discussion (in blog comments and twitter) about what women should do before they are allowed to join the conversation is ridiculous. Jeff s blog post says that men should listen to women, excluding women based on the tone argument is gross hypocrisy. Swearing Jeff makes a big deal of the fact that Shanley uses some profane language in her tweets. This combines a couple of different ways of silencing women. It s quite common for women to be held to a high standard of ladylike behavior, while men get a free pass on doing the same thing. One example of this is the Geek Feminism article about the results of Sarah Sharp s request for civility in the Linux kernel community [11]. That s not an isolated incident, to the best of my recollection in 20+ years my local Linux Users Group has had only one debate about profanity on mailing lists in that case a woman (who is no longer active in the group) was criticised for using lesser profanity than men used both before and after with no comment (as an experiment I used some gratuitous profanity a couple of weeks later and no-one commented). There is also a common difference in interpretation of expressions of emotion, when a woman seems angry then she invariably has men tell her to change her approach (even when there are obvious reasons for her anger) while when a man is angry the possibility that other people shouldn t make him angry will usually be considered. The issues related to the treatment of women have had a large affect on Shanley s life and her friend s lives. It s quite understandable that she is angry about this. Her use of profanity in tweets seems appropriate to the situation. Other Links Newsweek s Gentlemen in Technology article has a section about Jeff [12], it s interesting to note his history of deleting tweets and editing his post. I presume he will change his post in response to mine and not make any note of the differences. Jacob Kaplan-Moss wrote a good rebuttal to Jeff s post [13]. It s a good article and has some other relevant links that are worth reading.

The IT security world is still reeling from the impact of the OpenSSL Heartbleed bug. Thanks to the bug, many experts have been reviewing other technologies to try and find similar risks. While Heartbleed was hidden away in the depths of the OpenSSL code base, another major security risk has been hiding in plain sight: SMS authentication for web site logins. Remarkably, a number of firms have started giving customers the ability to receive single-use passwords over SMS for logging into their secure web sites. Some have even insisted that customers can no longer log in without it, denying customers the right to make an important choice about their own security preferences.

Unfortunately, SMS is no substitute to the one-time-passwords generated using proper authentication tokens or the use of other strong authentication schemes such as cryptographic smart cards. Even telephone companies themselves advise that SMS should not be used to secure financial transactions. Ocean's 11 in real life: exploiting the weakest link in the chain To deliver single-use SMS passwords, the SMS must travel through various networks from the firm's headquarters, to a wholesale SMS gateway, international SMS network and finally down the line of the local phone company. In comparison, properly certified token devices generate a code inside the device in the palm of your hand. The code only travels from the screen to your eyes. In a litany of frauds coming in all shapes and sizes, telephone networks have been exploited over and over again because they are almost always the weakest link in the chain. Using the mobile SMS network for authentication is not building on solid ground - some experts even feel it is downright stupidity. One of the most serious examples was the theft of $150,000,000 from a pension fund deposited with JP Morgan: it was described as a real-life case of Ocean's 11. The authentication was meant to be a phone call rather than an SMS: a phone company employee who was in on the scam duly ensured the call never reached the correct place.

The insecurity of traditional telephone networks has been on display for all the world to see in the ongoing trial of News Corporation executives for phone hacking. If journalists from a tabloid newspaper can allegedly hack a dozen phones before their first cigarette of the day, is it really wise to use an insecure technology like SMS as the cornerstone of a security system for authorizing transactions? A fraud recently played out on many credit card holders in the UK exploited a low-tech feature of the phone system to trick people to believe they were safe by "calling back" to their bank. A plethora of new attack vectors

The staggering reality of the situation is that attackers don't even have to directly hack their victim's phones to access SMS messages. As the Android API documentation demonstrates, SMS reception is notified to all apps in real-time. Apps can process the messages even when the phone is sleeping and the message is not read by the user. Just consider all the apps on a phone that have requested permission to read incoming messages. There was an uproar recently when a new version of the Facebook app started demanding permissions to read incoming SMS. The app can't be installed if the user doesn't agree to these new permissions. WhatsApp, another popular app that has SMS access rights, was recently exposed in a major security scandal which revealed they use a phone's IMEI number as the password. When people install an app like Tinder (which does not yet request SMS access) is the security of their bank account likely to be at the front of their mind? Even if Facebook intends no harm, they have opened the floodgates by further de-sensitizing users to the risks of giving apps un-necessary access to their data. These companies are looking for every piece of data that could give them an edge in their customer profiling and marketing programs. Having real-time access to your SMS is a powerful way for them to understand your activities and feelings at every moment in the day. To facilitate these data analysis techniques, replicating and archiving your messages into their cloud databases (whether you can see them there or not) is par for the course. The cloud, of course, has become a virtual smorgasboard for cyber-criminals, including both hackers and occasionally insiders wanting to peek at private data or harvest it en-masse. Social networking and communication sites are built on a philosophy of sharing data to create interaction and excitement. Unfortunately, this is orthogonal to the needs of security. In this context, the telephone network itself may no longer be the weakest link in the chain. The diligent attacker only needs to look for the cloud operator with an unplugged security hole and use their system as a stepping stone to read any SMS they want, when they want. Would you notice a stray SMS? Maybe you feel that you would notice a stray SMS carrying a login code for your bank account. Would you always be able to react faster than the criminal however? Thanks to social networks, or location data inadvertently leaked by other apps the attacker can easily work out whether you are on holiday, at the gym, at a party or sleeping or in some other situation where you are not likely to check messages immediately. If you receive a flood of SMS spam messages (deliberately sent by an attacker) in the middle of the night and you put your phone into silent mode and ignore it, you may well miss one message that was a login to your bank account. SMS technology was never designed for secure activities. The inconvenience of SMS While security is a headline issue these days, it is also worth reflecting on the inconvenience of SMS in some situations. Travel is at the top of the list: SMS doesn't work universally when abroad. These are usually the times when the only way to access the bank is through the web site. After dealing with the irritations of the hotel or airport wifi registration, do you really need more stress from your bank's systems too? For some networks, SMS can be delayed by hours or days, sometimes never arriving at all. Many people swap their SIM cards when travelling to avoid the excessive roaming charges and there is extra inconvenience in swapping SIM cards back again just to log in to a bank account. Worst of all, if you are tethering with a SIM card from the country you are visiting, then it is impossible for you to receive the SMS message from the bank on your regular SIM card while simultaneously maintaining the SSL connection to their web site over your new SIM card. Other problems like a flat battery, water damage or PIN permanently blocked by children playing with the phone can also leave you without access to your bank account for varying lengths of time. Is there any up-side to SMS authentication? The only potential benefit to SMS authentication is that it weeds out some of the most amateur attempts to compromise your bank account, but this is a false sense of security and it opens up new attack vectors through the cloud as we have just demonstrated. For all other purposes, it smells like a new form of security theater. A more likely reason why it has become popular amongst some firms is that many lenders want to ensure they have mobile phone numbers to contact customers when loan or credit card payments are missed. Making the mobile phone number mandatory for login ensures they almost always have the correct phone number for almost 100% of customers. It is not clear that this benefit justifies the failure to provide proper security and the inconvenience when travelling though. Opting out

Next time you log in to a web site, if the firm does try to enrol you in an SMS authentication scheme, it may be a good idea to click the "No thanks" option. If you have already been registered into an SMS authentication scheme, fill out the online complaint form and inform the firm that you will only accept a proper authentication token or cryptographic smart card. These solutions are tried and tested and they are the correct tool for the job.

My alternative option to quitting my day job so I could spend more time with Zoe was to take three months leave without pay. That wasn't really a great option, because it would mean that Zoe would only get one term of Kindergarten, and then I'd be back at work. Well today is the day that I'd go back to work if I took 3 months leave without pay. So far, I have zero regrets from doing the stay year-long at home Dad thing. At the end of the first week, I was absolutely exhausted, and a little worried that I'd bitten off more than I could chew, but since that point, I've found my stride and now I'm mostly just worried about what I'm going to do with myself next year. I'm really happy with the range of activities that I've been able to do with Zoe in the past 90 days. For this term, I'm going to do a combination of repeating some of the things we've done previously (Zoe's already asking to go ice skating again), and introduce her to a few new things that we didn't get a chance to do last term. As it starts to cool down a little bit more, we'll be looking around for more indoor activities to do, as well. Today though, I had a bit of a motivational slump. Zoe woke up, briefly, at 4:30am. I can't even remmeber why. I was smack bang in the middle of some serious REM sleep, so I found getting up to attend to her particularly jarring. I was a bit sluggish this morning as a result, and Zoe was also doing some spectacular procrastination. That coupled with rain meant we were a bit late getting to Kindergarten. The drop off itself went pretty well, and I drove home and had a late shower. There was a myriad of things I could have been doing, but I mostly just made a bunch of phone calls to organise things, did some more planning for Zoe's birthday party this Saturday, and waited for pick up time to roll around. I'll need to do a better job of being self-disciplined tomorrow. Kindergarten pick up was fine. I've been encouraging Zoe to reconsider napping at rest time, and the last two times I've picked her up, she hasn't napped, which has made for a much easier departure, and she hasn't been noticeably worse for wear for it. We went to take a look at the park we're having her birthday party in on Saturday, just to get a lie of the land in terms of tables and BBQs, and then we went to the Valley to clear my PO box. After that we went to the dance store in the Valley to get a leotard, socks and stockings for her ballet class. By the time we got home from all of that, there was only about an hour before Sarah was due to pick Zoe up, so she watched a bit of TV, and then we played hide and seek for a bit, and watered the seeds we've planted. I skipped yoga tonight because I'm still on lifting restrictions, and decided to skip Meatless Monday as well and hang out with my friend Michael and eat all the ribs I could handle. Boy was that an eating binge.

Distributing software efficiently accross your platform can be difficult. Every distribution comes with a package manager which is usually suited for this task. APT can be relied upon on when using Debian or a derivative. Unfortunately, the official repositories may not contain everything you need. When you require unpackaged software or more recent versions, it is possible to setup your own local repository. Most of what is presented here was setup for Dailymotion and was greatly inspired by the work done by Rapha l Pinson at Orange.

Setting up your repositories There are three kinds of repositories you may want to setup:

A distribution mirror. Such a mirror will save bandwidth, provide faster downloads and permanent access, even when someone searches Google on Google.
A local repository for your own packages with the ability to have a staging zone to test packages on some servers before putting them in production.
Mirrors for unofficial repositories, like Ubuntu PPA. To avoid unexpected changes, such a repository will also get a staging and a production zone.

Before going further, it is quite important to understand what a repository is. Let s illustrate with the following line from my /etc/apt/sources.list:

deb http://ftp.debian.org/debian/ unstable main contrib non-free

In this example, http://ftp.debian.org/debian/ is the repository and unstable is the distribution. A distribution is subdivided into components. We have three components: main, contrib and non-free. To setup repositories, we will use reprepro. This is not the only solution but it has a good balance between versatility and simplicity. reprepro can only handle one repository. So, the first choice is about how you will split your packages in repositories, distributions and components. Here is what matters:

A repository cannot contain two identical packages (same name, same version, same architecture).
Inside a component, you can only have one version of a package.
Usually, a distribution is a subset of the versions while a component is a subset of the packages. For example, in Debian, with the distribution unstable, you choose to get the most recent versions while with the component main, you choose to get DFSG-free software only.

If you go for several repositories, you will have to handle several reprepro instances and won t be able to easily copy packages from one place to another. At Dailymotion, we put everything in the same repository but it would also be perfectly valid to have three repositories:

one to mirror the distribution,
one for your local packages, and
one to mirror unofficial repositories.

Here is our target setup: Local APT repository

Initial setup First, create a system user to work with the repositories:

$ adduser --system --disabled-password --disabled-login \
>         --home /srv/packages \
>         --group reprepro

All operations should be done with this user only. If you want to setup several repositories, create a directory for each of them. Each repository has those subdirectories:

conf/ contains the configuration files,
gpg/ contains the GPG stuff to sign the repository¹,
logs/ contains the logs,
www/ contains the repository that should be exported by the web server.

Here is the content of conf/options:

outdir +b/www
logdir +b/logs
gnupghome +b/gpg

Then, you need to create the GPG key to sign the repository:

$ GNUPGHOME=gpg gpg --gen-key
Please select what kind of key you want:
   (1) RSA and RSA (default)
   (2) DSA and Elgamal
   (3) DSA (sign only)
   (4) RSA (sign only)
Your selection? 1
RSA keys may be between 1024 and 4096 bits long.
What keysize do you want? (2048) 4096
Requested keysize is 4096 bits
Please specify how long the key should be valid.
         0 = key does not expire
      <n>  = key expires in n days
      <n>w = key expires in n weeks
      <n>m = key expires in n months
      <n>y = key expires in n years
Key is valid for? (0) 10y
Key expires at mer. 08 nov. 2023 22:30:58 CET
Is this correct? (y/N) y
Real name: Dailymotion Archive Automatic Signing Key
Email address: the-it-operations@dailymotion.com
Comment: 
[...]

By setting an empty password, you allow reprepro to run unattended. You will have to distribute the public key of your new repository to let APT check the archive signature. An easy way is to ship it in some package.

Local mirror of an official distribution Let s start by mirroring a distribution. We want a local mirror of Ubuntu Precise. For this, we need to do two things:

Setup a new distribution in conf/distributions.
Configure the update sources in conf/updates.

Let s add this block to conf/distributions:

# Ubuntu Precise
Origin: Ubuntu
Label: Ubuntu
Suite: precise
Version: 12.04
Codename: precise
Architectures: i386 amd64
Components: main restricted universe multiverse
UDebComponents: main restricted universe multiverse
Description: Ubuntu Precise 12.04 (with updates and security)
Contents: .gz .bz2
UDebIndices: Packages Release . .gz
Tracking: minimal
Update: - ubuntu-precise ubuntu-precise-updates ubuntu-precise-security
SignWith: yes

This defines the precise distribution in our repository. It contains four components: main, restricted, universe and multiverse (like the regular distribution in official repositories). The Update line starts with a dash. This means reprepro will mark everything as deleted before updating with the provided sources. Old packages will not be kept when they are removed from Ubuntu. In conf/updates, we define the sources:

# Ubuntu Precise
Name: ubuntu-precise
Method: http://fr.archive.ubuntu.com/ubuntu
Fallback: http://de.archive.ubuntu.com/ubuntu
Suite: precise
Components: main main multiverse restricted universe
UDebComponents: main restricted universe multiverse
Architectures: amd64 i386
VerifyRelease: 437D05B5
GetInRelease: no
# Ubuntu Precise Updates
Name: ubuntu-precise-updates
Method: http://fr.archive.ubuntu.com/ubuntu
Fallback: http://de.archive.ubuntu.com/ubuntu
Suite: precise-updates
Components: main restricted universe multiverse
UDebComponents: main restricted universe multiverse
Architectures: amd64 i386
VerifyRelease: 437D05B5
GetInRelease: no
# Ubuntu Precise Security
Name: ubuntu-precise-security
Method: http://fr.archive.ubuntu.com/ubuntu
Fallback: http://de.archive.ubuntu.com/ubuntu
Suite: precise-security
Components: main restricted universe multiverse
UDebComponents: main restricted universe multiverse
Architectures: amd64 i386
VerifyRelease: 437D05B5
GetInRelease: no

The VerifyRelease lines are GPG key fingerprint to use to check the remote repository. The key needs to be imported in the local keyring:

$ gpg --keyring /usr/share/keyrings/ubuntu-archive-keyring.gpg \
>     --export 437D05B5   GNUPGHOME=gpg gpg --import

Another important point is that we merge three distributions (precise, precise-updates and precise-security) into a single distribution (precise) in our local repository. This may cause some difficulties with tools expecting the three distributions to be available (like the Debian Installer²). Next, you can run reprepro and ask it to update your local mirror:

$ reprepro update

This will take some time on the first run. You can execute this command every night. reprepro is not the fastest mirror solution but it is easy to setup, flexible and reliable.

Repository for local packages Let s configure the repository to accept local packages. For each official distribution (like precise), we will configure two distributions:

precise-staging contains packages that have not been fully tested and not ready to go to production.
precise-prod contains production packages copied from precise-staging.

In our workflow, packages are introduced in precise-staging where they can be tested and will be copied to precise-prod when we want them to be available for production. You can adopt a more complex workflow if you need. The reprepro part is quite easy. We add the following blocks into conf/distributions:

# Dailymotion Precise packages (staging)
Origin: Dailymotion #  
Label: dm-staging   #  
Suite: precise-staging
Codename: precise-staging
Architectures: i386 amd64 source
Components: main role/dns role/database role/web #  
Description: Dailymotion Precise staging repository
Contents: .gz .bz2
Tracking: keep
SignWith: yes
NotAutomatic: yes #  
Log: packages.dm-precise-staging.log
 --type=dsc email-changes
# Dailymotion Precise packages (prod)
Origin: Dailymotion #  
Label: dm-prod      #  
Suite: precise-prod
Codename: precise-prod
Architectures: i386 amd64 source
Components: main role/dns role/database role/web #  
Description: Dailymotion Precise prod repository
Contents: .gz .bz2
Tracking: keep
SignWith: yes
Log: packages.dm-precise-prod.log

First notice we use several components (in ):

main will contain packages that are not specific to a subset of the platform. If you put a package in main, it should work correctly on any host.
role/* are components dedicated to a subset of the platform. For example, in role/dns, we ship a custom version of BIND.

The staging distribution has the NotAutomatic flag (in ) which disallows the package manager to install those packages except if the user explicitely requests it. Just below, when a new dsc file is uploaded, the hook email-changes will be executed. It should be in the conf/ directory. The Origin and Label lines (in ) are quite important to be able to define an explicit policy of which packages should be installed. Let s say we use the following /etc/apt/sources.list file:

# Ubuntu packages
deb http://packages.dm.gg/dailymotion precise main restricted universe multiverse
# Dailymotion packages
deb http://packages.dm.gg/dailymotion precise-prod    main role/dns
deb http://packages.dm.gg/dailymotion precise-staging main role/dns

All servers have the precise-staging distribution. We must ensure we won t install those packages by mistake. The NotAutomatic flag is one possible safe-guard. We also use a tailored /etc/apt/preferences:

Explanation: Dailymotion packages of a specific component should be more preferred
Package: *
Pin: release o=Dailymotion, l=dm-prod, c=role/*
Pin-Priority: 950
Explanation: Dailymotion packages should be preferred
Package: *
Pin: release o=Dailymotion, l=dm-prod
Pin-Priority: 900
Explanation: staging should never be preferred
Package: *
Pin: release o=Dailymotion, l=dm-staging
Pin-Priority: -100

By default, packages will have a priority of 500. By setting a priority of -100 to the staging distribution, we ensure the packages cannot be installed at all. This is stronger than NotAutomatic which sets the priority to 1. When a package exists in Ubuntu and in our local repository, we ensure that, if this is a production package, we will use ours by using a priority of 900 (or 950 if we match a specific role component). Have a look at the How APT Interprets Priorities section of apt_preferences(5) manual page for additional information. Keep in mind that version matters only when the priority is the same. To check if everything works as you expect, use apt-cache policy:

$ apt-cache policy php5-memcache
  Installed: 3.0.8-1~precise2~dm1
  Candidate: 3.0.8-1~precise2~dm1
  Version table:
 *** 3.0.8-1~precise2~dm1 0
        950 http://packages.dm.gg/dailymotion/ precise-prod/role/web amd64 Packages
        100 /var/lib/dpkg/status
     3.0.8-1~precise1~dm4 0
        900 http://packages.dm.gg/dailymotion/ precise-prod/main amd64 Packages
       -100 http://packages.dm.gg/dailymotion/ precise-staging/main amd64 Packages
     3.0.6-1 0
        500 http://packages.dm.gg/dailymotion/ precise/universe amd64 Packages

If we want to install a package from the staging distribution, we can use apt-get with the -t precise-staging option to raise the priority of this distribution to 990. Once you have tested your package, you can copy it from the staging distribution to the production distribution:

$ reprepro -C main copysrc precise-prod precise-staging wackadoodle

Local mirror of third-party repositories Sometimes, you want a software published on some third-party repository without to repackage it yourself. A common example is the repositories edited by hardware vendors. Like for an Ubuntu mirror, there are two steps: defining the distribution and defining the source. We chose to put such mirrors into the same distributions as our local packages but with a dedicated component for each mirror. This way, those third-party packages will share the same workflow as our local packages: they will appear in the staging distribution, we validate them and copy them to the production distribution. The first step is to add the components and an appropriate Update line to conf/distributions:

Origin: Dailymotion
Label: dm-staging
Suite: precise-staging
Components: main role/dns role/database role/web vendor/hp
Update: hp
# [...]
Origin: Dailymotion
Label: dm-prod
Suite: precise-prod
Components: main role/dns role/database role/web vendor/hp
# [...]

We added the vendor/hp component to both the staging and the production distributions. However, only the staging distribution gets an Update line (remember, packages will be copied manually into the production distribution). We declare the source in conf/updates:

# HP repository
Name: hp
Method: http://downloads.linux.hp.com/SDR/downloads/ManagementComponentPack/
Suite: precise/current
Components: non-free>vendor/hp
Architectures: i386 amd64
VerifyRelease: 2689B887
GetInRelease: no

Don t forget to add the GPG key to your local keyring. Notice an interesting feature of reprepro: we copy the remote non-free component to our local vendor/hp component. Then, you can synchronize the mirror with reprepro update. Once the packages have been tested, you will have to copy them in the production distribution.

` directory.

You can propose your package for inclusion into Debian. This will help many people once the package hits the archive.

Builders We chose pbuilder as a builder⁴. Its setup is quite straightforward. Here is our /etc/pbuilderrc:

DISTRIBUTION=$DIST
NAME="$DIST-$ARCH"
MIRRORSITE=http://packages.dm.gg/dailymotion
COMPONENTS=("main" "restricted" "universe" "multiverse")
OTHERMIRROR="deb http://packages.dm.gg/dailymotion $ DIST -staging main"
HOOKDIR=/etc/pbuilder/hooks.d
BASE=/var/cache/pbuilder/dailymotion
BASETGZ=$BASE/$NAME/base.tgz
BUILDRESULT=$BASE/$NAME/results/
APTCACHE=$BASE/$NAME/aptcache/
DEBBUILDOPTS="-sa"
KEYRING="/usr/share/keyrings/dailymotion-archive.keyring.gpg"
DEBOOTSTRAPOPTS=("--arch" "$ARCH" "--variant=buildd" "$ DEBOOTSTRAPOPTS[@] " "--keyring=$KEYRING")
APTKEYRINGS=("$KEYRING")
EXTRAPACKAGES=("dailymotion-archive-keyring")

pbuilder is expected to be invoked with DIST, ARCH and optionally ROLE environment variables. Building the initial bases can be done like this:

for ARCH in i386 amd64; do
  for DIST in precise; do
    export ARCH
    export DIST
    pbuilder --create
  done
done

We don t create a base for each role. Instead, we use a D hook to add the appropriate source:

#!/bin/bash
[ -z "$ROLE" ]    
  cat >> /etc/apt/sources.list <<EOF
deb http://packages.dm.gg/dailymotion $ DIST -staging role/$ ROLE 
EOF
 
apt-get update

We ensure packages from our staging distribution are preferred over other packages by adding an /etc/apt/preferences file in a E hook:

#!/bin/bash
cat > /etc/apt/preferences <<EOF
Explanation: Dailymotion packages are of higher priority
Package: *
Pin: release o=Dailymotion
Pin-Priority: 900
EOF

We also use a C hook to get a shell in case there is an error. This is convenient to debug a problem:

#!/bin/bash
apt-get install -y --force-yes vim less
cd /tmp/buildd/*/debian/..
/bin/bash < /dev/tty > /dev/tty 2> /dev/tty

A manual build can be run with:

$ ARCH=amd64 DIST=precise ROLE=web pbuilder \
>         --build somepackage.dsc

Version numbering To avoid to apply complex rules to chose a version number for a package, we chose to treat everything as a backport, even in-house software. We use the following scheme: `X-Y~preciseZ+dmW`.

`X` is the upstream version⁵.

`Y` is the Debian version. If there is no Debian version, use 0.

`Z` is the Ubuntu backport version. Again, if such a version doesn t exist, use 0.

`W` is our version of the package. We increment it when we make a change to the packaging. This is the only number we are allowed to control. All the others are set by an upstream entity, unless it doesn t exist and in this case, you use 0.

Let s suppose you need to backport `wackadoodle`. It is available in a more recent version of Ubuntu as `1.4-3`. Your first backport will be `1.4-3~precise0+dm1`. After a change to the packaging, the version will be `1.4-3~precise0+dm2`. A new upstream version `1.5` is available and you need it. You will use `1.5-0~precise0+dm1`. Later, this new upstream version will be available in some version of Ubuntu as `1.5-3ubuntu1`. You will rebase your changes on this version and get `1.5-3ubuntu1~precise0+dm1`. When using Debian instead of Ubuntu, a compatible convention could be : `X-Y~bpo70+Z~dm+W`.

Uploading To upload a package, a common setup is the following workflow:

Upload the source package to an incoming directory.
reprepro will notice the source package, check its correctness (signature, distribution) and put it in the archive.
The builder will notice a new package needs to be built and build it.
Once the package is built, the builder will upload the result to the incoming directory.
reprepro will notice again the new binary package and integrate it in the archive.

This workflow has the disadvantage to have many moving pieces and to leave the user in the dark while the compilation is in progress. As an alternative, a simple script can be used to execute each step synchronously. The user can follow on their terminal that everything works as expected. Once we have the .changes file, the build script just issues the appropriate command to include the result in the archive:

$ reprepro -C main include precise-staging \
>      wackadoodle_1.4-3~precise0+dm4_amd64.changes

Happy hacking!

The gpg/ directory could be shared by several repositories.
We teached Debian Installer to work with our setup with an appropriate preseed file.
fpm-cookery is a convenient tool to write recipes for fpm, similar to Homebrew or a BSD port tree. It could be used to achieve the same goal.
sbuild is an alternative to pbuilder and is the official builder for both Debian and Ubuntu. Historically, pbuilder was more focused on developers needs.
For a Git snapshot, we use something like 1.4-git20130905+1-ae42dc1 which is a snapshot made after version 1.4 (use 0.0 if no version has ever been released) at the given date. The following 1 is to be able to package different snapshots at the same date while the hash is here in case you need to retrieve the exact snapshot.

Current Glamor Performance I finally managed to get a Gigabyte Brix set up running Debian so that I could do some more reasonable performance characterization of Glamor in its current state. I wanted to use this particular machine because it has enough cooling to keep from thermally throttling the CPU/GPU package. This is running my glamor-server branch of the X server, which completes the core operation rework and then has some core X server performance improvements as well for filled and outlined arcs. Changes in X11perf First off, I did some analysis of what x11perf was doing and found that it wasn t quite measuring what we thought. I m not interested in competing on x11perf numbers absolutely, I m only interested in doing relative measurements of useful operations, so fixing the tool to measure what I want seems reasonable to me. When x11perf was first written, it drew 100x100 rectangles tight against one another without any gap. And, it filled the window with them, drawing a 6x6 grid of 100x100 rectangles in a 600x600 window. To better exercise the rectangle code and check edge conditions better, we added a one pixel gap between the rectangles. However, we didn t reduce the number of rectangles drawn, so we ended up drawing 11 of the 36 rectangles on top of the first set of 25. Simple region computations would allow the X server to draw only 25 most of the time, skipping the redundant rectangles. The vertical and horizontal line tests were added a while after the first set of tests, and were done without regard to how an X server might optimize for them. x11perf draws these lines packed tightly together, creating a single square of pixels for the result. EXA, UXA and SNA all take vertical and horizontal lines and convert them to rectangles, then take the rectangles and clip them against the window clip list by computing a region from them and intersecting that with the GC composite clip. It s a completely reasonable plan, however, when you take what x11perf was drawing and run it through this code, you end up with a single solid rectangle. Which is surprisingly fast, compared with drawing individual lines. I fixed the overlapping rectangle case by reducing the number of boxes drawn from 36 to 25, and I fixed the vertical and horizontal line case by spacing the lines a pixel apart. I ve pushed out these changes to my x11perf repository on freedesktop.org. What s Fast Things that match GL s capabilities are fast, things which don t are slow. No surprises there. What s interesting is precisely what matches GL Patterns For Free Because GL makes it easy to program fill patterns into the GPU, there are essentially no performance differences between solid and patterned operations. GL Lines Glamor uses GL lines, which can be programmed to match X semantics, to quite good effect. The only trick required was to deal with cap styles. GL never draws the final pixel in a line, while X does unless the cap style is CapNotLast. The solution was to draw an extra line segment containing a single pixel at the end of every joined set of lines for this case. The other implicit requirement is that all zero width lines look the same. Right now, I ve solved that for fill styles and raster ops as they re all drawn with the same GL operations. However, for plane masks, we re currently falling back to software, which may draw something different. Fixing that isn t impossible, it s just tedious. Text Pushing all of the work of drawing core text into glamor wasn t terribly difficult, and the results are pretty spectacular. What s Slow We ve still got room for improvement in Glamor, but there aren t any obvious show-stoppers to getting great performance for reasonable X applications anymore. Wide Lines and Arcs One of the speed-ups I ve made in my glamor branch is to merge all of drawing of multiple filled and zero-width arcs into a single giant GL request. That turned out to both improve performance and save a bit of code. Right now, drawing wide lines and wide arcs doesn t do this, and so we suffer from submitting many smaller requests to GL. It s hard to get excited about speeding any of this up as all of the wide primitives are essentially unused these days. Filled Polygons Because X only lets applications draw a single polygon in each request, Glamor can t really gain any efficiency from batching work unless we start looking ahead in the X protocol stream to see if the next request is another polygon. Alternatively, we could leave the span operation pending to see if more spans were coming before the X serve went idle. Neither of these is all that exciting though; X polygons just aren t that useful. Render Operations These are still not structured to best fit modern GL; some work here would help a bunch. We ve got a gsoc student ready to go at this though, so I expect we ll have much better numbers in a few months. Window Operations You wouldn t think that moving and resizing windows would be so limited by drawing performance, but x11perf tests these with tiny little windows, and each operation draws or copies only a couple of little rectangles, which makes GL quite expensive. Working on speeding up GL for small numbers of operations would help a bunch here. Unexpected Results Solid rectangles are actually running slower than patterned rectangles, and I really have no idea why. The CPU is close to idle during the 500x500 solid rectangle test (as you d expect, given the workload), the vertex and fragment shaders look correct out of the compiler, and yet solid rectangles run at only 0.80 of the performance of the patterned rectangles. GL semantics for copying data essentially preclude overlapping blts of any form. There s the NVtexturebarrier extension which at least lets us do blts within the same object, but even that doesn t define how overlapping blts work. So, we have to create a temporary copy for this operation to make it work right. I was worried that this would slow things down, but the Iris Pro 3D engine is enough faster than the 2D engine that even with the extra copy, large scrolls and copies within the same object are actually faster. Results Here s a giant image showing the ratio of Glamor to both UXA and SNA running on the same machine, with all of the same software; the only change between runs was to switch the configured acceleration architecture.

I ve seen many comments about swap space and SSD claiming that swap will inherently destroy SSD through using too many writes. The latest was in the comments of my post about swap space and SSD performance [1]. Note that I m not criticising the person who commented on my blog, everyone has heard lots of reports about possible problems that they avoid without analysing them in detail. The first thing to note is that the quality of flash memory varies a lot, the chips that are used in SSDs for workstation/server use are designed to last while those in USB-flash devices aren t. I ve documented my unsuccessful experiments with using USB-flash for the root filesystem of a gateway server [2] (and the flash device that wasn t used for swap died too). The real issue when determining whether swap will break your SSD is the amount of writes. While swapping can do a lot of writing quickly that usually doesn t happen unless something has gone wrong. The workstation that I m currently using has writes to the root fileystem outnumbering writes to swap by a factor of 130:1 (by volume of data written). On other days I ve seen it as low as 42:1, in either case it s writes to the root filesystem (which is BTRFS and includes /home etc) that will break the SSD if anything. For some other workstations I run I see ratios of 201:1 (that s with 8G of RAM), 57:1, and 23:1. In a quick search I couldn t find a single system I run where even 10% of disk writes were attributed to swap. This really isn t surprising given that adding RAM is a cheap way to improve the performance of most systems. If a SSD didn t do any write leveling (as is rumored to be the case with cheap USB flash devices) then swap use might still cause a problem because of the number of writes in a small area, but if that was the case then filesystem journals and other fixed data structures would be more likely to cause a problem and any swap based breakage would break swap not the root filesystem (although if swap was the first partition then it might also break the MBR). Of the workstations that are convenient to inspect the one with the most writes to the root filesystem and swap space (IE everything on /dev/sda) had 128G written per day for 1.2 days of uptime, but that involved running a filesystem balance and some torrent downloads (not the typical use). Among the two systems which had been running for more than 48 hours with typical use the most writes was 24G in a day. If a SSD can sustain 10,000 writes per block (which is smaller than quoted by most flash manufacturers nowadays) and has perfect wear leveling (which is unlikely) then the system with a 120G SSD and 128G written in a day could continue like that for almost 10,000 days or 27 years much longer than storage is expected to work or be useful. So for workstation use (where even 24G of writes per day probably counts as heavy use) a 120G SSD that can sustain 10,000 writes per block shouldn t be at great risk of wearing out. Conclusion I believe that swap is much less likely to break SSDs than regular file access on every system with a reasonable amount of RAM. For systems which don t have enough RAM you probably want the speed of SSD for swap space anyway in spite of the risks. If wear leveling works as designed and the 10,000+ writes per block claims are accurate then SSDs will massively outlast their useful life. Long before they wear out they should be too small, too slow, and probably not compatible. 26 years ago I had a 5.25 full height ST-506 disk in my desktop PC, it wouldn t physically fit in most systems I own now (unless I removed the DVD drive), there is no possibility of buying a controller (I don t own a system with an ISA bus), and that disk was too slow and small by today s standards. A 27yo SSD isn t going to be useful for anything, even for archive storage it s no good as no-one has tested long term storage and data could decay before then.

Since my blog post about BTRFS in March [1] not much has changed for me. Until yesterday I was using 2.13 kernels on all my systems and dealing with the occasional kmail index file corruption problem. Yesterday my main workstation ran out of disk space and went read-only. I started a BTRFS balance which didn t seem to be doing any good because most of the space was actually in use so I deleted a bunch of snapshots. Then my X session aborted (some problem with KDE or the X server I ll never know as logs couldn t be written to disk). I rebooted the system and had kernel threads go into infinite loops with repeated messages about a lack of response for 22 seconds (I should have photographed the screen). When it got into that state the ALT-Fn keys to change a virtual console sometimes worked but nothing else worked the terminal usually didn t respond to input. To try and stop the kernel from entering an infinite loop on every boot that I used rootflags=skip_balance on the kernel command line to stop it from continuing the balance which made the system usable for a little longer, unfortunately the skip_balance mount option doesn t permanently apply, the kernel will keep trying to balance the filesystem on every mount until a btrfs balance cancel operation succeeds. But my attempts to cancel the balance always failed. When I booted my system with skip_balance it would sometimes free some space from the deleted snapshots, after two good runs I got to 17G free. But after that every time I rebooted it would report another Gig or two free (according to btrfs filesystem df ) and then hang without committing the changes to disk. I solved this problem by upgrading my USB rescue image to kernel 3.14 from Debian/Experimental and mounting the filesystem from the rescue image. After letting kernel 3.14 work on the filesystem for a while it was in a stage where I could use it with kernel 3.13 and then boot the system normally to upgrade it to kernel 3.14. I had a minor extra complication due to the fact that I was running apt-get dist-upgrade at the time the filesystem went read-only do the dpkg records of which packages were installed were a bit messed up. But that was easy to fix by running a diff against /var/lib/dpkg/info on a recent snapshot. In retrospect I should have copied from an old snapshot of the root filesystem, but I fixed the problems faster than I could think of better ways to fix them. When running a balance the system had a peak IO rate of about 30MB/s reads and 30MB/s writes. That compares to the maximum contiguous file IO speed of 260MB/s for reads and 320MB/s for writes. During that time it had about 50% CPU time used for my Q8400 quad-core CPU. So far the only tasks that I do regularly which have CPU speed as a significant bottleneck are BTRFS filesystem balancing and recoding MP4 files. Compiling hasn t been an issue because recently I haven t been compiling many programs that are particularly big. Lessons Learned I should photograph the screen regularly when doing things that won t be logged, those kernel error messages might have been useful to me or someone else. The fact that the only kernel that runs BTRFS the way I need comes from the Experimental repository in Debian stands in contrast to the recent kernel patch that stops describing BTRFS as experimental. While I have a high opinion of the people who provide support for the kernel in commercial distributions and their ability to back-port fixes from newer kernels I m concerned about their decision to support BTRFS. I m also dubious about whether we can offer BTRFS support in Debian/Jessie (the next version of Debian) without a significant warning. OTOH if you find yourself with a BTRFS system that isn t working well you could always hire me to fix it. I accept payment via Paypal, bank transfer, or Bitcoin. If you want to pay me in Grange then I assure you I will never forget about it. ;) I thought that I wouldn t have CPU speed issues when I started using the AMD64 architecture, for most tasks that s been the case. But for systems for which storage is important I ll look at getting faster CPUs because of BTRFS. Using faster CPUs for storage isn t that uncommon (I used to work for SGI and dealt with some significant CPU power used for file serving), but needing a fast quad-core CPU to drive a single SSD is a little disappointing. While recovery from file system corner cases isn t going to be particularly common it s something that you want completed quickly, for personal systems you want to be doing something else and for work systems you don t want down-time. The BTRFS problems with running out of disk space are really serious. It seems that even workstations used at home can t survive without monitoring. For any other filesystem used at home you can just let it get full and then delete stuff. Include rootflags=skip_balance in the boot loader configuration for every system with a BTRFS root filesystem and in the /etc/fstab for every non-root BTRFS filesystem. I haven t yet encountered a single situation where continuing the balance did any good or when it didn t do any harm.

[1] http://etbe.coker.com.au/2014/03/16/btrfs-march-2014/

The Debian Project is pleased to announce that registration is now open for DebConf14, taking place in Portland, Oregon, USA from Saturday, August 23 to Sunday, August 31, 2014. To request food, accommodation, or travel sponsorship, you must be registered by Thursday, May 15, 2014. After this date, registrations will still be accepted in any of the basic, professional, and corporate categories, but requests for sponsorship will no longer be accepted. Note that, as for previous years, you will be asked to confirm your registration later once sponsorship has been finalized. If you are not certain you will attend but are considering it, we encourage you to register now. Registration If you want to attend DebConf14, please fill out the registration form. This year, conference registration is integrated with the new Debian SSO system. If you are a Debian developer and have not previously used the Debian SSO system, you will need to configure an SSO password on db.debian.org. If you are not a Debian developer, you will have an opportunity to create an alioth account as part of the registration process. For more information about Debian single sign-on, see the Debian wiki. DebCamp In past years, DebConf has been preceded by a separate DebCamp event, for developers to gather before the conference and collaborate in person on Debian. This year the team is trying a different approach, with a longer conference period allowing for blocks of talks to be mixed with blocks of coding time throughout the week. More information about the schedule will be made available when the Call for Papers is posted in the near future. As mentioned in the last bits from the DPL mail, there is also an opportunity for Debian teams to organize sprints around DebConf. Teams who are interested should follow the documented process. Thank you The DebConf14 team would like to thank all of our sponsors who have made this event possible - a full list of whom can be found here. We are still seeking sponsors to help us make DebConf14 a success. If you or your company would like to give back to Debian, please consider becoming a sponsor. We look forward to seeing you in Portland in August! The DebConf team

Some people seem to be justifying the way mobile networks try to discriminate against tethering. One comment even suggests that multiple devices is a valid justification. Really? Isn't it just as simple as measuring the actual volume of megabytes used, regardless of how many devices are connected? Lets consider a couple of analogies:

A petrol station puts some fine print at the bottom of their sign in the street: cars with passengers pay an extra 0.05 per litre (or gallon). It is just a gimmick: the marketing man has worked out that he can put a lower price on his big sign this way and people with passengers may well be splitting the fuel bill between them. There is no technical reason for the charge however - the car actually uses more fuel carrying the extra weight anyway.
Or imagine this: the marketing man puts up small print that advises a higher charge of 0.05 per litre for fuelling an exotic car like a Ferrari. The only reason for this, of course, is that the Ferrari driver is probably busy (so he doesn't have time to shop around) and wealthy (so the extra charge won't deter him from making any purchase at all, because he has money to pay it).

Of course, these ideas sound a bit outlandish but the only difference between these crazy ideas and tethering charges is that mobile phone networks usually have the monopoly/duopoly to get away with it. Mobile networks actually dread the idea of becoming commodities like fuel. They thrive on price discrimination and they seem to have an endless number of tricks to get extra money out of people who may have it by deliberately crippling parts of the service for everybody else. This goes against the ideals of innovation and efficiency and in the most extreme cases their practices are widely recognised for the sham that they are. Revealing that you have connected a laptop to your phone may be like revealing you live in an affluent area or that you have an expensive car. Plenty of businesses try to exploit such knowledge when deciding how much they can charge you or other aspects of how they provide you with a service. This is why privacy is important in every feature of the technology we use. This rogue behaviour by "creative" marketing men also helps to understand just what would happen in the world of technology if it wasn't for the use of free licenses, open standards, accessible encryption and net neutrality.

After quite a few hours of work, I finally switched completely to DNSSEC. Both client-side in my notebook, and in my personal tincho.org domain. The client-side was pretty easy, although something broke in dnsmasq, but I had no patience to debug it, so I have just replaced it with a stock bind9 install, which is DNSSEC-enabled by default nowadays! To complement that, I added a plugin to Firefox/Iceweasel (WNPP bug #672845 pending, downloadable from the Czech NIC) that shows me with a nice icon if the DNS is secure or not (and in the newest versions, it also shows DANE status, yay!). So, basically, if you want to have DNSSEC support in your computer, just install unbound or bind9 (maybe dnsmasq, if you don't hit the same bug as me), it is really easy to have it up and running in no time. To test if it is correctly working, apart from that nifty plugin, you can visit this funny web page from Verisign labs. On the server side, it was trickier. It involves quite a few steps, and the default tools from package dnssec-tools are pretty buggy. But it was not too bad. After moving my domain from the registrar's DNS to my own server, configuring secondaries, etc. I went ahead with the DNSSEC configuration. I used this pretty good DNSSEC howto, which made the process a lot easier. After having my DNS server ready, I added the DS records in my registrar, and voil , tincho.org is now protected by DNSSEC! There are a few web services to test your deployment, a simple one, and a more complete one with GraphViz diagrams! I felt so bold with all this, that I went ahead and created DANE and SSHFP records for my services (and had to debug issues with SSH, because the old ssh-keygen tool would not create ECDSA records). And even set Postfix to use DANE to connect to remote hosts. Let's see how many things break in the following days!

On the last UDS we talked about migrating from upstart to systemd to boot Ubuntu, after Mark announced that Ubuntu will follow Debian in that regard. There s a lot of work to do, but it parallelizes well once developers can run systemd on their workstations or in VMs easily and the system boots up enough to still be able to work with it. So today I merged our systemd package with Debian again, dropped the systemd-services split (which wasn t accepted by Debian and will be unnecessary now), and put it into my systemd PPA. Quite surprisingly, this booted a fresh 14.04 VM pretty much right away (of course there s no Plymouth prettiness). The main two things which were missing were NetworkManager and lightdm, as these don t have an init.d script at all (NM) or it isn t enabled (lightdm). Thus the PPA also contains updated packages for these two which provide a proper systemd unit. With that, the desktop is pretty much fully working, except for some details like cron not running. I didn t go through /etc/init/*.conf with a small comb yet to check which upstart jobs need to be ported, that s now part of the TODO list. So, if you want to help with that, or just test and tell us what s wrong, take the plunge. In a 14.04 VM (or real machine if you feel adventurous), do

  sudo add-apt-repository ppa:pitti/systemd
  sudo apt-get update
  sudo apt-get dist-upgrade

This will replace systemd-services with systemd, update network-manager and lightdm, and a few libraries. Up to now, when you reboot you ll still get good old upstart. To actually boot with systemd, press Shift during boot to get the grub menu, edit the Ubuntu stanza, and append this to the linux line: init=/lib/systemd/systemd. For the record, if pressing shift doesn t work for you (too fast, VM, or similar), enable the grub menu with

  sudo sed -i '/GRUB_HIDDEN_TIMEOUT/ s/^/#/' /etc/default/grub
  sudo update-grub

Once you are satisfied that your system boots well enough, you can make this permanent by adding the init= option to /etc/default/grub (and possibly remove the comment sign from the GRUB_HIDDEN_TIMEOUT lines) and run sudo update-grub again. To go back to upstart, just edit the file again, remove the init=sudo update-grub again.


I ll be on the Debian systemd/GNOME sprint next weekend, so I feel reasonably well prepared now.  
Update: As the comments pointed out, this bricked /etc/resolv.conf. I now uploaded a resolvconf package to the PPA which provides the missing unit (counterpart to the /etc/init/resolvconf.conf upstart job) and this now works fine. If you are in that situation, please boot with upstart, and do the following to clean up:
  sudo rm /etc/resolv.conf
  sudo ln -s ../run/resolvconf/resolv.conf /etc/resolv.conf

Then you can boot back to systemd.
Update 2: If you want to help testing, please file bugs with a systemd-boot tag. See the list of known bugs when booting with systemd.

Outlier detection (also: anomaly detection, change detection) is an unsupervised data mining task that tries to identify the unexpected.

Most outlier detection methods are based on some notion of density: in an appropriate data representation, "normal" data is expected to cluster, and outliers are expected to be further away from the normal data.

This intuition can be quantified in different ways. Common heuristics include kNN outlier detection and the Local Outlier Factor (which uses a density quotient). One of the directions in my dissertation was to understand (also from a statistical point of view) how the output and the formal structure of these methods can be best understood.

I will present two smaller results of this analysis at the SIAM Data Mining 2014 conference: instead of the very heuristic density estimation found in above methods, we design a method (using the same generalized pattern) that uses a best-practise from statistics: Kernel Density Estimation. We aren't the first to attempt this (c.f. LDF), but we actuall retain the properties of the kernel, whereas the authors of LDF tried to mimic the LOF method too closely, and this way damaged the kernel.

The other result presented in this work is the need to customize. When working with real data, using "library algorithms" will more often than not fail. The reason is that real data isn't as nicely behaved - it's dirty, it seldom is normal distributed. And the problem that we're trying to solve is often much narrower. For best results, we need to integrate our preexisting knowledge of the data into the algorithm. Sometimes we can do so by preprocessing and feature transformation. But sometimes, we can also customize the algorithm easily.

Outlier detection algorithms aren't black magic, or carefully adjusted. They follow a rather simple logic, and this means that we can easily take only parts of these methods, and adjust them as necessary for our problem at hand!

The article persented at SDM will demonstrate such a use case: analyzing 1.2 million traffic accidents in the UK (from data.gov.uk) we are not interested in "classic" density based outliers - this would be a rare traffic accident on a small road somewhere in Scotland. Instead, we're interested in unusual concentrations of traffic accidents, i.e. blackspots.

The generalized pattern can be easily customized for this task. While this data does not allow automatic evaluation, many outliers could be easily verified using Google Earth and search: often, historic imagery on Google Earth showed that the road layout was changed, or that there are many news reports about the dangerous road. The data can also be nicely visualized, and I'd like to share these examples with you. First, here is a screenshot from Google Earth for one of the hotspots (Cherry Lane Roundabout, North of Heathrow airport, which used to be a double cut-through roundabout - one of the cut-throughs was removed since):

Screenshot of Cherry Lane Roundabout hotspot

Google Earth is best for exploring this result, because you can hide and show the density overlay to see the crossroad below; and you can go back in time to access historic imagery. Unfortunately, KML does not allow easy interactions (at least it didn't last time I checked).

I have also put the KML file on Google Drive. It will automatically display it on Google Maps (nice feature of Drive, kudos to Google!), but it should also allow you to download it. I've also explored the data on an Android tablet (but I don't think you can hide elements there, or access historic imagery as in the desktop application).

With a classic outlier detection method, this analysis would not have been possible. However, it was easy to customize the method; and the results are actually more meaningful: instead of relying on some heuristic to choose kernel bandwidth, I opted for choosing the bandwidth by physical arguments: 50 meters is a reasonable bandwidth for a crossroad / roundabout, and for comparison a radius of 2 kilometers is used to model the typical accident density in this region (there should other crossroads within 2 km in Europe).

Since I advocate reproducible science, the source code of the basic method will be in the next ELKI release. For the customization case studies, I plan to share them as a how-to or tutorial type of document in the ELKI wiki; probably also detailing data preprocessing and visualization aspects. The code for the customizations is not really suited for direct inclusion in the ELKI framework, but can serve as an example for advanced usage.

Reference:
E. Schubert, A. Zimek, H.-P. Kriegel
Generalized Outlier Detection with Flexible Kernel Density Estimates
In Proceedings of the 14th SIAM International Conference on Data Mining (SDM), Philadelphia, PA, 2014.

So TLDR of the story: A) try to use more established statistics (such as KDE), and B) don't expect an off-the-shelf solution to do magic, but customize the method for your problem.

P.S. if you happen to know nice post-doc positions in academia:
I'm actively looking for a position to continue my research. I'm working on scaling these methods to larger data and to make them work with various real data that I can find. Open-source, modular and efficient implementations are very important to me, and one of the directions I'd like to investigate is porting these methods to a distributed setting, for example using Spark. In order to get closer to "real" data, I've started to make these approaches work e.g. on textual data, mixed type data, multimedia etc. And of course, I like teaching; which is why I would prefer a position in academia.

White Trash

The above poster was on a bridge pylon in Flinders St in 2012. It s interesting to see what the Fringe Festival people consider to be associated with white trash . They claim homophobia is a white trash thing however lower class people have little political power and the fact that we still don t have marriage equality in Australia is clear evidence that homophobia is prevalent among powerful people. Toys vs Fairies

Fairies look pretty while boys toys do things

I took the above photo at Costco in 2012. I think it s worth noting the way that the Disney Fairies (all female and marketed to a female audience) are standing around looking pretty while the Toy Story characters (mostly male and marketed to a male audience) are running out to do things. Having those items side by side on the shelf was a clear example of a trend in toys towards girls being encouraged to be passive while boys are doing things. The Toy Story pack has one female character, so it could be interpreted as being aimed at both boys and girls. But even that interpretation doesn t remove the clear gender difference. It seems ironic to me that the descriptions on the boxes are Read, Play, and Listen for the Toy Story pack and Read, Play, and Colour on the Fairies pack. Colouring is more active than listening so the pictures don t match the contents. Make Up vs Tools

Girls chocolate is make-up and boys chocolate is tools

I took the above photo in an Aldi store in early 2013, today I was in Aldi and noticed that the same chocolate is still on sale. A clear and pointless gender difference. Rumor has it that some of the gender difference in kids clothing is so that a child can t wear the clothes of an older sibling of different gender, but chocolate only gets eaten once so there is no reason for this. Oath

The above poster was inside the male toilet at Melbourne University in 2013. It would probably be good to have something like that on display all the time instead of just for one event. Locks

Locks with inscriptions on a bridge on the Yarra River in Melbourne

I took the above picture early this year, it shows hundreds of padlocks attached to a bridge across the Yarra River in Melbourne. Each padlock has a message written or inscribed in it, mostly declarations of love. I first noticed this last year, I m not sure how long it s been up. There was nothing formal about this (no signs about it), people just see it and decide that they want to add to it. I guess that the council cuts some of them off periodically as the number of locks doesn t seem to be increasing much in recent times. It would be interesting to do some research into how many locks are needed to start one of these. It would also be interesting to discover whether the nature of the inscriptions determines the speed at which it takes off, would a bunch of padlocks with messages like I Love Linux inspire others as well as messages declaring love for random people? All that is required is some old locks and an engraving tool. I wonder what the social norm might be regarding messing with those locks. If I was to use those padlocks to practice the sport of lock-picking (which I learned when in Amsterdam) I wonder whether random bystanders would try to discourage me. It seems likely that picking the locks and taking them away would get a negative reaction but I wonder whether picking them one at a time and replacing them (or maybe moving them to another wire) would get a reaction. Blackface for Schoolkids

teachers choice blackface and yellowface masks

A craft shop at the Highpoint shopping center in Melbourne is selling Teacher s Choice brand Multicultural Face Masks . Multicultural is a well regarded term in education, teaching children about other cultures is a good concept but can be implemented really badly. When I was in high school the subject Social Studies seemed to have an approach of look how weird people are in other places instead of teaching the kids anything useful. Sociological Images has an informative article on the Australian Hey Hey it s Saturday blackface incident in 2009 [1]. The idea of these masks seems to involve students dressing up as caricatures of other races. The mask which looks like someone s idea of a Geisha is an even bigger WTF, mixing what the package calls culture (really race) with sex work. When I visited Tokyo I got the impression that French maids fill a similar niche to Geisha for younger Japanese men and the maid cafe thing is really popular there. I think it s interesting to consider the way that a French maid costume is regarded differently to a Geisha costume. I expect that Teacher s Choice doesn t sell French maid costumes. Delicious Cow

Usually meat is advertised in a way that minimises the connection to living animals. Often adverts just show cuts of meat and don t make any mention of animals and when animals are shown they are in the distance. The above picture was on the wall at a Grill d burger restaurant in Point Cook. It shows a bovine (looks like a bull even though I believe that cows are the ones that are usually eaten) with a name-tag identifying it as Delicious . The name tag personalises the animal which is an uncommon thing to do when parts of an animal are going to be eaten. Of the animals that are commonly eaten it seems that the general trend is to only show fish as complete live animals, presumably because people can identify with mammals such as cattle in a way that they can t identify with fish. Fish are also the only complete animals that are shown dead, adverts for fish that are sold as parts (EG salmon and tuna) often show complete dead fish. But I ve never seen a meat advert that shows a complete dead cow or sheep.

[1] http://tinyurl.com/jvvamgb

I picked up a Panasonic BDT-230 a couple of months ago. Then I discovered that even though it appeared fairly straightforward to make it DVD region free (I have a large pile of PAL region 2 DVDs), the US models refuse to play back PAL content. We live in an era of software-defined functionality. While Panasonic could have designed a separate hardware SKU with a hard block on PAL output, that would seem like unnecessary expense. So, playing with the firmware seemed like a reasonable start.

Panasonic provide a nice download site for firmware updates, so I grabbed the most recent and set to work. Binwalk found a squashfs filesystem, which was a good sign. Less good was the block at the end of the firmware with "RSA" written around it in large letters. The simple approach of hacking the firmware, building a new image and flashing it to the device didn't appear likely to work.

Which left dealing with the installed software. The BDT-230 is based on a Mediatek chipset, and like most (all?) Mediatek systems runs a large binary called "bdpprog" that spawns about eleventy billion threads and does pretty much everything. Runnings strings over that showed, well, rather a lot, but most promisingly included a reference to "/mnt/sda1/vudu/vudu.sh". Other references to /mnt/sda1 made it pretty clear that it was the mount point for USB mass storage. There were a couple of other constraints that had to be satisfied, but soon attempting to run Vudu was actually setting a blank root password and launching telnetd.

/acfg/config_file_global.txt was the next stop. This is a set of tokens and values with useful looking names like "IDX_GB_PTT_COUNTRYCODE". I tried changing the values, but unfortunately made a poor guess - on next reboot, the player had reset itself to DVD region 5, Blu Ray region C and was talking to me in Russian. More inconveniently, the Vudu icon had vanished and I couldn't launch a shell any more.

But where there's one obvious mechanism for running arbitrary code, there's probably another. /usr/local/bin/browser.sh contained the wonderful line:

export LD_PRELOAD=/mnt/sda1/bbb/libSegFault.so

, so then it was just a matter of building a library that hooked open() and launched inetd and dropping that into the right place, and then opening the browser.

This time I set the country code correctly, rebooted and now I can actually watch Monkey Dust again. Hurrah! But, at the same time, concerning. This software has been written without any concern for security, and it listens on the network by default. If it took me this little time to find two entirely independent ways to run arbitrary code on the device, it doesn't seem like a stretch to believe that there are probably other vulnerabilities that can be exploited with less need for physical access.

The depressing part of this is that there's no reason to believe that Panasonic are especially bad here - especially since a large number of vendors are shipping much the same Mediatek code, and so probably have similar (if not identical) issues. The future is made up of network-connected appliances that are using your electricity to mine somebody else's Dogecoin. Our nightmarish dystopia may be stranger than expected.

comments

In 2011 I wrote a post that was inspired by the Sociological Images blog [1]. After some delay here I ve written another one. I plan to continue documenting such things. Playground

In 2011 I photographed a plaque at Flagstaff Gardens in Melbourne. It shows a picture of the playground in 1918 with segregated boys and girls sections. It s interesting that the only difference between the two sections is that the boys have horizontal bars and a trapeze. Do they still have gender segregated playgrounds anywhere in Australia? If so what is the difference in the sections? Aborigines The Android game Paradise Island [2] has a feature where you are supposed to stop Aborigines from stealing, it plays on the old racist stereotypes about Aborigines which are used to hide the historical record that it s always been white people stealing from the people that they colonise.

There is also another picture showing the grass skirts. Nowadays the vast majority of Aborigines don t wear such clothing, the only time they do is when doing some sort of historical presentation for tourists. I took those pictures in 2012, but apparently the game hasn t changed much since then. Lemonade

Is lemonade a drink or a flavour? Most people at the party where I took the above photo regard lemonade as a drink and found the phrase Lemonade Flavoured Soft Drink strange when it was pointed out to them. Incidentally the drink on the right tastes a bit like the US version of lemonade (which is quite different from the Australian version). For US readers, the convention in Australia is that lemonade has no flavor of lemons. Not Sweet

In 2012 an apple cider company made a huge advertising campaign featuring people who might be gender queer, above is a picture of a bus stop poster and there were also TV ads. The adverts gave no information at all about what the drink might taste like apart from not being as sweet as you think . So it s basically an advertising campaign with no substance other than a joke about people who don t conform to gender norms. Also it should be noted that some women naturally grow beards and have religious reasons for not shaving [3]. Episode 2 of the TV documentary series Am I Normal has an interesting interview of a woman with a beard. Revolution

A violent political revolution is usually a bad thing, using such revolutions to advertise sugar drinks seems like a bad idea. But it seems particularly interesting to note the different attitudes to such things in various countries. In 2012 Schweppes in Australia ran a marketing campaign based on imagery related to a Communist revolution (the above photo was taken at Southern Cross station in Melbourne), I presume that Schweppes in the US didn t run that campaign. I wonder whether global media will stop such things, presumably that campaign has the potential to do more harm in the US than good in Australia. Racist Penis Size Joke at Southbank

The above advert was in a free newspaper at Southbank in 2012. Mini Movers thought that this advert was a good idea and so did the management of Southbank who approved the advert for their paper. Australia is so racist that people don t even realise they are being racist.

Early this month at a LUV meeting I gave a talk with only my mobile phone to store notes. I used Google Keep to write the notes as it s one of the easiest ways of writing a note on a PC and quickly transferring it to a phone if I keep doing this I will find some suitable free software for this task. Owncloud seems promising [1], but at the moment I m more concerned with people issues than software. Over the years I ve experimented with different ways of presenting lectures. I m now working with the theory that presenting the same data twice (by speaking and text on a projector) distracts the audience and decreases learning. Editing and Viewing Notes Google Keep is adequate for maintaining notes, it s based on notes that are a list of items (like a shopping list) which is fine for lecture notes. It probably has lots of other functionality but I don t care much about that. Keep is really fast at updating notes, I can commit a change on my laptop and have it visible on my phone in a few seconds over 3G. Most of the lectures that I ve given have involved notes on a laptop. My first laptop was a Thinkpad 385XD with a 12.1 display and all my subsequent laptops have had a bigger screen. When a laptop with a 12 or larger screen is on a lectern I can see the notes at a glance without having to lean forward when 15 or fewer lines of text are displayed on the screen. 15 lines of text is about the maximum that can be displayed on a slide for the audience to read and with the width of a computer display or projector is enough for a reasonable quantity of text. When I run Keep on my Galaxy Note 2 it displays about 20 rather short lines of text in a portrait orientation (5 points for a lecture) and 11 slightly longer lines in a landscape orientation (4 points). In both cases the amount of text displayed on a screen is less than that with a laptop while the font is a lot smaller. My aim is to use free software for everything, so when I replace Keep with Owncloud (or something similar) I will probably have some options for changing the font size. But that means having less than 5 points displayed on screen at a time and thus a change in the way I present my talks (I generally change the order of points based on how well the audience seem to get the concepts so seeing multiple points on screen at the same time is a benefit). The Samsung Galaxy Note 2 has a 5.5 display which is one of the largest displays available in a phone. The Sony Xperia X Ultra is one of the few larger phones with a 6.44 display that s a large phone but still not nearly large enough to have more than a few points on screen with a font readable by someone with average vision while it rests on a lectern. The most obvious solution to the problem of text size is to use a tablet. Modern 10 tablets have resolutions ranging from 1920*1080 to 2560*1600 and should be more readable than the Thinkpad I used in 1998 which had a 12 800*600 display. Another possibility that I m considering is using an old phone, a Samsung Galaxy S weighs 118 to 155 grams and is easier to hold up than a Galaxy Note 2 which weighs 180g. While 60g doesn t seem like much difference if I m going to hold a phone in front of me for most of an hour the smaller and lighter phone will be easier and maybe less distracting for the audience. Distributing URLs When I give a talk I often want to share the addresses of relevant web sites with the audience. When I give a talk with the traditional style lecture notes I just put the URLs on the final page (sometimes using tinyurl.com) for people to copy during question time. When I use a phone I have to find another way. I did a test with QR code recognition and found that a code that takes up most of the width of the screen of my Galaxy Note 2 can be recognised by a Galaxy S at a distance of 50cm. If I ran the same software on a 10 tablet then it would probably be readable at a distance of a meter, if I had the QR code take up the entire screen on a tablet it might be readable at 1.5m away, so it doesn t seem plausible to hold up a tablet and allow even the first few rows of the audience to decode a QR code. Even if newer phones have better photographic capabilities than the Galaxy S that I had available for testing there are still lots of people using old phones who I want to support. I think that if QR codes are to be used they have to be usable by at least the first three rows of the audience for a small audience of maybe 50 people as that would allow everyone who s interested to quickly get in range and scan the code at the end. Chris Samuel has a photo (taken at the same meeting) showing how a QR code from a phone could be distributed to a room [2]. But that won t work for all rooms. One option is to just have the QR code on my phone and allow audience members to scan it after the lecture. As most members of the audience won t want the URLs it should be possible for the interested people to queue up to scan the QR code(s). Another possibility I m considering is to use a temporary post on my documents blog (which isn t syndicated) for URLs. The WordPress client for Android works reasonably well so I could edit the URL list at any time. That would work reasonably well for talks that have lots of URLs which is quite rare for me. A final option is to use Twitter, at the end of a talk I could just tweet the URLs with suitable descriptions. A good portion of the Tweets that I have written is URLs for web sites that I find interesting so this isn t a change. This is probably the easiest option, but with the usual caveat of using a proprietary service as an interim measure until I get a free software alternative working. Any suggestions? Please comment if you have any ideas about ways of addressing these issues. Also please let me know if anyone is working on a distributed Twitter replacement. Please note that anything which doesn t support followers on multiple servers and re-tweets and tweeting to users on other servers isn t useful in this regard.

[1] https://owncloud.com/
[2] http://tinyurl.com/nww3wu4

Debian 7.5 will include an update to the Linux kernel, based on Linux 3.2.57. Package version 3.2.57-2 is currently available in the wheezy-proposed-updates suite. I would appreciate any testing people can do to find regressions in the next few days. In addition to bug fixes, this version updates the e1000e and igb drivers. The drivers are now based on the versions found in Linux 3.13, which support several newer chips (i210, i211, i217, i218, i354). Please consider testing this new kernel if you have an Intel gigabit Ethernet controller, even if it was already supported in Linux 3.2.

I gave a talk this year at PyCon 2014, about one of my favorite subjects: Hy. Many of my regular readers will have no doubt explored Hy's thriving GitHub org, played with try-hy, or even installed it locally by pip installing it. I was lucky enough to be able to attend PyCon on behalf of Sunlight, with a solid contingint of my colleagues. We put together a writeup on the Sunlight blog if anyone was interested in our favorite talks.

Tons of really amazing questions, and such an amazingly warm reception from so many of my peers throughout this year's PyCon. Thank you so much to everyone that attended the talk. As always, you should Fork Hy on GitHub, follow @hylang on the twitters, and send in any bugs you find! Hopefully I'll be able to put my talk up in blog-post form soon, but until then feel free to look over the slides or just watch the talk. An extra shout-out to @akaptur for hacking on Hy during the sprints, and giving the exception system quite the workthrough. Thanks, Allison!

Search Results: "reg"

30 April 2014

Wouter Verhelst: Code of Conduct accepted

29 April 2014

Gunnar Wolf: Split regarding Docker

Russell Coker: Autism and the Treatment of Women Again

28 April 2014

Daniel Pocock: SMS logins: an illusion of security

Andrew Pollock: [life] Day 90: Kindergarten, and a motivational slump

27 April 2014

Vincent Bernat: Local corporate APT repositories

Keith Packard: Glamor performance

Russell Coker: Swap Breaking SSD

26 April 2014

Russell Coker: BTRFS Status April 2014

DebConf team: Registration is now open for DebConf14 (Posted by Steve Langasek)

25 April 2014

Daniel Pocock: Extra fuel charges when carrying passengers in your car?

Martín Ferrari: More DNSSEC

24 April 2014

Martin Pitt: Booting Ubuntu with systemd: Test packages available

22 April 2014

Erich Schubert: Kernel-density based outlier detection and the need for customization

21 April 2014

Russell Coker: Sociological Images 2014

Matthew Garrett: Home entertainment implementations are pretty appalling

20 April 2014

Russell Coker: Sociological Images 2012

19 April 2014

Russell Coker: Phone Based Lectures

Ben Hutchings: Linux kernel update for Debian 7.5; new Intel Ethernet drivers

Paul Tagliamonte: Hy at PyCon 2014