I've been recently investigating a poorly maintained box (granted, it was just a workstation used for surfing the web) which was behaving oddly. In fact you could not longer login via SSH. I quickly noticed that there was a non-working sshd on it, in /usr/local/bin. So I thought - why would a badly maintained box have a non-standard SSH on it? This made me very suspicious, so I immedeately ran chkrootkit. It didn't find anything except it told me that someone had tampered with the wtmp file. The "tampering dates" aligned with the creation times of the ssh. A quick "find" run came up with the tool to do so, too. A quick check in the logfiles - which were fine, since they were not standard syslog, probably - showed that a SSH scanner had managed to login into an unprivileged account with a weak password, then used a kernel exploit from january to gain root privileges. Apparently he was unsatisfied with the sshd on the box, so he tried to put a different version on it, which, well, didn't work and he could no longer logon to the box himself. A rootkit was apparently not yet installed (verified after a clean boot), he was just about to setup his own sshd... maybe he had noticed that the box was just a stupid surfing box and then didn't care enough to cover his tracks (or just shoot himself in the foot by breaking the sshd) So if you are running boxes on the internet:

• Make sure you don't allow shell logins with weak passwords
• Have your system and kernel up to date (note: you need to reboot for activating a new kernel...)
• Don't run an sshd unless you really need to
• Maybe setup a recent-match filter to stop ssh scanners
• Scan your log files for unusual entries such as logins by users who shouldn't really exist in the first place... Use logcheck.
• Don't allow PHP scripts to run binaries on your system ...
• Or better, don't allow PHP at all

(PHP guys: yes, I know that there are some good PHP scripts. But there are tons of badly written ones, too... and PHP is a major intrusion vector, and the privilege escalation I've seen here would have worked just fine with a PHP installation not using safe mode and safe_mode_exec_dir) A firewall I maintain at our university has been tracking ssh connections using the recent match for quite some time, and a nice side effect is that it reduces all their "spam" in your logs, too (in case you bother to read them. Do at least read the logcheck results!)

WARNING: nothing in this post has been implemented or tested, except maybe at a minimal proof-of-concept level ("it compiles!"). It's just some musings on how you can make the computer do your work for you.

I was reading a defense of Hungarian notation when I started to get a familiar notion tickling the back of my brain. The author was explaining that the original idea behind Hungarian was not to replicate type information in variable names, but rather to include some sort of usage information: one example would be giving a string value holding a value that is untrusted the name utName. While it's more reasonable than lpschwzMyVariable, this suggesion made me suspicious that the author was really hacking around suckiness in the C language.

Tagging values with usWhatever and then verifying that you never mix usBlah with Blah isn't just tedious and error-prone, it smells a lot like doing work by hand that the computer ought to do for you. It seemed to me that it ought to be possible to "tag" unsafe values at the type level and let the language's type-checker sort things out. The simple approach is of course to define something like this:

newtype Unsafe t = Unsafe t

NB: I'm using Haskell because it was the first language that sprang to mind when I thought "type system hack".

This newtype declaration lets you attach an Unsafe tag to any type. Presumably the type signatures of routines that return untrusted data (such as a hypothetical getHTMLFormValue or a routine to retrieve previously entered form values from a database, if they aren't sanitized first) should be modified; for instance, if the signature of getHTMLFormValue was

getHTMLFormValue :: HTMLForm -> String -> String

it should instead read

getHTMLFormValue :: HTMLForm -> String -> Unsafe String

To create a new Unsafe value you simply apply the Unsafe constructor to the value:

getHTMLFormValue = Unsafe (oldGetHTMLFormValue)

To unwrap Unsafes you use pattern-matching:

let (Unsafe unwrapped) = wrapped in ...

So, this sort of works, but it's unsatisfying on several levels:

• It allows the creation of types such as Unsafe Unsafe Integer and even Unsafe [Unsafe Integer]. If a value is already unsafe, how can it be "more unsafe"? And what in the world is an unsafe list, or an unsafe list of unsafe values?
  
  Unfortunately, this is probably a limitation that any scheme that relies on type-tagging will have. You can get around it by either explicitly defining each new unsafe type:
  
  

  newtype UnsafeString = Unsafe String

  
  
  This is arguably a good idea anyway, since virtually all unsafe values (in the sense the previously linked article mentiones) are likely to be strings, and as I note later, it's not clear how to "sanitize" a non-string. OTOH, placing unnecessary type restrictions feels un-Haskellish to me, and it seems likely that you can trust the programmer not to pile up Unsafes.
  
  
• This scheme makes it very tempting to strip off the unsafeness of a value and lose the fact that you did so.
  
  Since the only way to strip off unsafeness is via pattern-matching, it seems to me that it would be tempting to pattern-match away the unsafeness in function declarations:
  
  

  foo (Unsafe x) = ...

  
  
  This may be just fine, but my preference would be to forbid pattern-matching altogether (by not exporting the constructor directly to other modules; just export a normal function that constructs an Unsafe value) and instead providing two routines to remove unsafety:
  
  

  escapeData :: Unsafe t -> t
  unsafeExtractRawData :: Unsafe t -> t

  
  
  The first routine encodes the data into a "safe" form, while the second simply fetches the wrapped value:
  
  

  unsafeExtractRawData (Unsafe x) = x
  
  foo x = doSomeStuffWith (unsafeExtractRawData x)

  
  
  The rationale here is twofold: first, by giving the dangerous extraction routine a scary and hard-to-type name, you remind yourself to think twice before invoking it (cf unsafePerformIO); second, you can easily audit all the locations where the unsafety is "stripped away" with a textual search.
  
  There are some issues involving the safe encoding -- you probably ought to make a typeclass of things that can be safely encoded from an Unsafe value; not too hard and this entry is already too long.
  
  
• The final issue involves manipulation of Unsafe values. So far you can pack and unpack Unsafes; however, this seems like it would get a bit annoying if you had to do any significant operations on them. For instance, searching in an unsafe string for a substring seems like it would be a pain and a bit error-prone (since you'd have to remember to wrap the string back up in an Unsafe). You might think that ideally, all normal String operations should work directly on an Unsafe String or pairs of Unsafe Strings -- you just shouldn't be able to convert between the two.
  
  Unfortunately, I can't see any way of doing everything that you'd want to here. For some types you can. For instance, the following definitions let you compare values wrapped in Unsafe in the usual way:
  
  

  instance (Eq t) => Eq (Unsafe t) where
      (==) (Unsafe v1) (Unsafe v2)   = v1 == v2
      (/=) (Unsafe v1) (Unsafe v2)   = v1 /= v2

  
  
  This admits (Unsafe t) to the Eq class, and looks reasonable enough (assuming that it's safe to compare unsafe values, of course). However, for probably the most important Unsafe values -- Unsafe Strings -- I can't see how to do this. The problem is that String is not a typeclass, it's a synonym for [Char], a list of characters. The important string operations are either defined directly on the String type, or generically on [t]; i.e., "a list of anything". Because there is no typeclass, you simply can't overload these operations to apply to unsafe strings as well as regular strings.
  
  One way around this is to provide an "unwrapping combinator":
  
  

  liftUnsafe :: (t -> t) -> Unsafe t -> Unsafe t
  liftUnsafe f (Unsafe v) = Unsafe (f v)

  
  
  This would automatically convert string operations to unsafe-string operations; not ideal, but reasonable enough. Operators are still a little annoying, but you can mangle their names:
  
  

  (++u) :: Unsafe [t] -> Unsafe [t] -> Unsafe [t]
  (++u) = liftUnsafe (++)

  
  
  So you have to use ++u instead of ++ to concatenate unsafe lists; not too bad.
  

While it may not be perfect -- it only makes sure that you don't accidentally violate safety, and doesn't stop you from deliberately doing something dumb -- this approach feels a lot safer to me than ad-hoc mangling of variable names. I used Haskell because you can do type hacks fairly easily in it, but I don't see why you couldn't play a similar game (albeit much less conveniently) in C++, or even in a dynamic language like Python. Of course, Python would give you runtime errors about safety violations instead of static ones, but that still beats getting runtime security holes.

I would add as a hypothesis that it should be possible to express just about any variable-name-mangling technique whose goal is to "make wrong code look wrong" in the type system, provided your type system doesn't suck; and that doing so is preferable to doing it manually, since it means that a lot of repetitive error-checking can be handled by the computer. And since the computer is better than you are at verifying formal invariants, forcing it to handle them whenever possible is always a good idea.

According to Jeff, I’ve become a smelly Nokia contractor (the reason he stayed with Luis instead of at Mako’s place in Boston, although I’m sure he was referring to Rob Taylor causing the smell). Part of this entails exchanging project schedules with various managers, where time is often talked about in terms of numbered weeks of the year, like: Foo task will be completed by week 42 of 2005. We ran into some problems with varying definitions of these numbers when exchanging schedules with people running on different platforms. Then we realised that even the software we’re using seems to have different ideas about what’s going on too…

• Outlook and Evolution seem to agree, and define the week as starting on Monday, and the first day of the year is always week 1, even if this results in a truncated week. Hence, week 1 of 2006 starts and ends on Sunday 1st January, and then week 2 begins on Monday 2nd, etc. It’s my belief that this is the correct behaviour as defined by some ISO standard somewhere which governs week numbers.
• GtkCalendar gets the week number correct, in that the first day of the year is always in week 1, but varies its definition of the first day of the week depending on your locale. This is presumably so that the display of the weeks is correct, but using it to number the weeks results in bogus week numbers. For me in LOCALE=en_GB it claims that week 1 of 2006 starts on Sunday 1st January, and ends on Saturday 7th January, making my week numbers one lower than the rest of the world.
• Planner uses GtkCalendar for inputting dates, so gets it wrong there as described, but seems to get it wrong in its Gantt chart view in a different way. It has the week starting on Monday, but claims that January 1st 2006 is the last day of week 52 of 2005. This results in week numbers that, match those GtkCalendar give me on weekdays, but differ over weekends, and are still totally bogus when I try and talk about project plans with my managers.

The result is a whole world of pain, and at best causes extreme confusion when we provide documents referring to both dates and week numbers which are inconsistent with each other in their minds, and at worst makes our managers think we’ll have things done a week sooner than we do. I’m not using Evolution at the moment for calendaring stuff, so I can manage if I remember to add one to all my 2006 week numbers when interacting with managers based on GtkCalendar and Planner, but this all seems to be horribly broken. I really don’t want to have to do this for the whole year.

Social network services (Wikipedia) such as Friendster, Linkedin, OpenBC are one of the top "web 2.0" hypes. I'm very sceptical about them, because I don't see the real benefits for me. (I'm referring to the pure "keep contact with your schoolmates" thing, not stuff which does much more such as serious blogging - e.g. livejournal - or planets such as planet.debian) The typical use case I've heard is much like "Imagine you have made a cool product, and want to sell it to some company. Now you can search for people at that company you know (or a friend of yours knows) and call them, this will help you get a feet in the door", as well as "you can use it for finding a job". I think that's bullshit, and to a large extend because I'd neither be willing to update my information in these services all the time, nor would I be willing to help just everybody here - most likely I'd just delete and ignore the email, and be really annoyed by a call (I'm annoyed by calls anyway)... It's not that I despise social networking per se. But I despise quantity instead of quality. I care a lot for my friends, and I'm in several groups grown around some technical issues; when someone is looking for a job I pay attention for suiteable offers around me; if someone needs something else I do try to help. But I'll only do that for people I'd really call friends. If you just want a quantitative number, last I checked my ranking was #36 in the global PGP keyring, meaning I have the 36th shortest average distance worldwide to verify a cryptographical key. So what? Of course I also tried one of the social networing services given above, joined a group for a university I was at. Even before the adminstrator of that group added me, I got an "invitation" by one "linkaholic". I rejected it, and even considered to report it as "spam" to the operator. I did indeed know that person, but all we had was a flamewar on a mailinglist, and I found him highly unsympatic. So why should I add him to my network? Where are the benefits for me (except that I'll maybe get an annoying call sometime in the future, when I could have forgotten about him)? No, in my opinion this social networking hype is built entirely on a hype, basically saying "look how well-connected I am, everybody knows me, and I know everybody", even when these relationships are only present on the computer, without anything worth being called a "relationship" behind. The flame-war I mentioned earlier was btw. on the topic of Resume/CV books. Which are even worse than this social networking thing, IMHO.

Well, there's nothing like posting a really rubbish program to your blog to find out the right way to do it. Firstly, Philipp Kern has come up with another, even shorter version than his version I posted in the update to my original Ruby post, this time blowing me away with the strangeness of explicit casts in Ruby. Apparently, casting an input stream to an array produces an array with one-line-per-element, as in this example:

 STDIN.to_a.sort_by rand .each l puts l 

As Philipp warns, however: "Ruby is full of fun and power; one must watch out that the scripts remain readable". For a fairly toy example, this solution is still fairly readable, however I can see hair growing on this thing if your program grows large. Next, Pierre-Charles David showed his qualifications for becoming a co-author of the next edition of the Pickaxe (AKA "Programming Ruby: The Pragmatic Programmers' Guide"), with his example and explanation. He's kindly given permission for me to reproduce his e-mail in it's entireity, since it's too good to chop up:

Here is the shortest and most readable version of your "randomize stdin" script I can think of:

  #!/usr/bin/env ruby
  print ARGF.readlines.sort_by   rand  
 

ARGF is a special variable representing either stdin or, in the case where file names are passed on the command line, the content of these files (think of Perl's <>). ARGF behaves like an IO object (although I think it is not one). In particular, it responds to the #readlines method which splits the receiver's content into lines and returns an array. In other words,

  lines = ARGF.readlines
 

is the idiomatic way of doing

  lines = []
  while line = gets() do
    lines << line
  end
 

in your example. The next part (sort_by) uses a new form of the sort method, which I think is not documented in the Pickaxe. It implements the Schwartzian transform technique. In essence, a random number is associated to each element of the array, which are then sorted based on this index. As a general style remark, note that although Ruby supports "while" and "for x in y" loops, they are rarely used, in favor of iterators and/or higher order methods. For example, to gather all the input lines in an array without using readlines, one might write

  lines = Array.new
  ARGF.each_line    l  lines << l  
 

or

  lines = ARGF.inject([])    lines, l  lines << l  
 

Hope it helps, and welcome to Ruby!

Pierre, your explanation definitely helped, and I hope it helps others too. The third person to help me out is Decklin Foster, through a post in his blog. What particularly caught me about his solution was the neat demonstration of another feature of Ruby which I think I will really learn to love: the ability to dynamically add methods to an existing class. "Aiee!" I hear all the purists scream -- and I can see the potential for severe misuse, too. But, since every good idea seems to get deeply perverted, we may as well have some real useful features to pervert while we're at it. Decklin has the clever idea of adding methods to the Array class that shuffle the elements of the array in a random fashion. His code is beautiful in it's simplicity:

 class Array
    def swap!(a, b)
        self[a], self[b] = self[b], self[a]
        self
    end
    def shuffle!
        each_index do  i 
            j = i + rand(length - i)
            swap!(i, j)
        end
        self
    end
 end

He then goes on to rewrite my bumbling example in probably the most comprehensible manner I've yet seen:

 STDIN.readlines.shuffle!.each    line  puts line  

Decklin writes, "Now that is cool.". I couldn't have put it better myself. My thanks to Philipp, Pierre-Charles, and Decklin, for helping me along the Ruby path. So, two new Ruby Rules for me to live by:

• Whenever you reach for a while, first take a good look around for any iterator methods which might apply.
• Never dismiss the power of defining your own methods on an existing class if it will improve the comprehension of your program.

Hmm, is it possible that Ruby is nothing more than Functional Programming by Stealth? Aiee!

Where do I start? It's been an eventful couple of days... The flight to LA was uneventful. Qantas is certainly orders of magnitude more pleasant to fly with than United, even though we were seated in a bulkhead area. I had a 13-month old baby seated next to me (no, Mikal was on the other side), and he was about as well behaved as you could possibly ask a 13-month old baby to behave. Mikal also behaved himself. I had the intention of trying to get something approximating 8 hours sleep, as we arrived at LAX at 7am local time, but I only managed about 4 hours. I did spend about 8 hours with a mask on and my brain in idle, so I arrived at LAX being mostly coherent. Entering the US via LAX was a much less traumatic experience than I was expecting. The E-3 visa was accepted fine, and I now have an I-94 stapled into my passport. Quarantine involved getting looked up and down by a guy in a booth, who waved me through. We had 3 hours to spare (or about 2 by the time we cleared all the immigration/customs queues) and then hopped on a 50 minute American Airlines flight to San Francisco, hired a car and drove to Santa Clara. The accommodation we're staying in for a month is very spiffy indeed. It's a largish gated community of units. Mine's got a ground floor entry, but everything else (except the garage) is upstairs. When Sarah gets here, we'll put up some photos. The weather is lovely as well. Our first outing was to the Social Security office to attempt to get social security numbers. This wasn't as successful as I'd hoped, as we were too freshly in the country to show up on their systems, so we left with letters saying we'd get one within four weeks, when our details had been verified. Yesterday evening, we went for a general recce of the surrounding area, and checked out a supermarket. That was less disconcerting than I expected, although I don't think it was a terribly huge supermarket. Having a GPS certainly raises one's confidence when navigating significantly. Last night I discovered that all the keys in my apartment were not created equal, and that the two loose in the lock box on the front door were the only two that opened the unit, and the two sets on key rings on the coffee table only opened the amenities and my mailbox. So I got to spend a bit of time sitting on my front doorstep until the nice man from Synergy Relocations came out to let me in (except his key didn't work either, so he got to make conversation with a legal alien while we waited for someone else to come out with a key). Today, we drove around to see a shopping mall, and the first set of shops we found happened to be around the corner from an enormous (by my standards) apartment complex, which had some signage implying that it had vacancies, so we decided to get the low down on how renting worked before we did our rental tours. Casey drove us around in a golf buggy and gave a tour of what two different floor plans looked liked. It seems that in this part, it's standard for an apartment to include the refrigerator, dishwasher, microwave, washing machine and clothes dryer, which is good to know. Let me just say that if this is what living in an apartment complex is like, it's more like living in a resort. The place had pools everywhere, tennis courts, two gyms, a community centre, it was just unreal. It was also cheaper than I was expecting. We then went to a massive shopping mall to do some window shopping. Discovered Wetzel's Pretzels. Hope they're not too fattening... Tomorrow, I'm going to avail myself to the complex's gym, and generally get ready for the week ahead. Future posts will be in a new category, which I won't be feeding to Planet Debian as not to bore people. I will inflict Planet Linux Australia readers with them until someone asks me to stop. If you'd like a custom feed that excludes my American adventures, please let me know. I wonder if adjusting Blosxom's idea of timezone buggers up existing posts and aggregators like Planet in general?

Dear Lazyweb,
,
Nothing is perfect, but on Linux you can at least improve it and tune it. I just ran into memory problems on my development box - I had tons of applications open, and since this is a laptop I prefer to have swap disabled. I then started to build bind9 for a sarge backport, which meant that a long sequence of C compiler processes were launched, each allocating some memory, working with it, terminating, next. Being really at the limit of the available memory, the system became unresponsive - and that behavious is really bad... I thought that sooner or later it will really run out of memory, and decided to watch some TV in the meantime. One hour later when I returned, it was still torturing my HD and still unresponsive. Apparently the compile process has been all the time just below the memory limit, causing the kernel to remove pages from memory and load them back from the disk later, which slowed everything down just horribly. So now I'm searching for a way to prevent such situations. The traditional ulimit approach won't help - no single process was responsible for using all the memory, but each one had its reasonable share. So dear lazyweb, is there a way to tell the kernel to reserve a share of the memory (lets say, 64 MB) for buffers and caches only? I can live with out-of-memory errors a bit early (and the just enable swap), but a massive slowdown as described above (where the kernel is apparently reading some data from the HD, then forgetting it again and replacing it with the next) must not happen that easily... Right now I fear that you could bring down any Linux system easily with a "slow" forkbomb where each forked process allocates 100 MB and starts writing some random data to its memory, forking a new child every 50 MB or so. Neither the OOM killer, ulimit nor traditional forkbomb detectors will be able to counter that, since already a couple of processes will cause the system to become completely IO-bound. Oh, and I have no idea how Windows would react on that either. So don't assume any OS is better, just because I say Linux isn't doing perfectly well here. If you have any tips for me, send me an email.

Flying home from Frankfurt yesterday via Chicago, as usual I randomly selected a shortish looking line in the immigration and customs hall. When it was my turn, the agent behind the counter saw the "business" box checked on my form and asked who I work for, which is pretty standard. He then asked me what I do for HP, but before I could answer, he said "No, wait. You've been through here before. I recognize your passport." I managed to retain my composure just enough to agree that he certainly could have seen my passport before. He smiled, stamped my form, and waved me through. My wife wonders why I was so surprised.

Continuing on from the SCC stuff, which I should probably get in the habit of calling the mirror split, then… At the same time as I’ve been trying to work out ways to fit the mirror split stuff into the AJ market concept, I’ve been pondering on and off what can be done to improve the security situation. My thinking’s probably still biassed by my long time wish for official security updates to testing, but I’m somewhat inclined to think that if we can figure out a way of integrating the testing security team and the official security team we might end up closer to a solution – at the very least that’d provide us with a decent number of people who can actively help out with security updates that aren’t secet and get a few more people familiar with the technical aspects of how security updates are released, who’ll then have less of a learning curve if they end up being added to the security team proper. And heck, even if that doesn’t work, it at least gives people running testing an easier url to remember for security updates. Of course, integrating the teams isn’t that easy, or it would’ve already been done. So after thinking about it, I had a chat with Joey (aka Martin Schulze, aka the guy who’s issued 317 of the 340 security advisories in the past 12 months) about the problems:

<aj> so, the other thing i wonder is what you think of the testing-security-support split?
<Joey> need more input
<aj> as in separate host, funny domain name, not available on security.debian.org/, extra effort to mirror, etc
<aj> i think it sucks :)
<Joey> there are a lot of sucking things…
<aj> it’s true
<aj> so, i figure the reasons for the split are (a) you don’t want to add a bunch of people to team@ and hence vendor-sec, and (b) there’s no way to make it so they can use dak-security without being full team/vendor-sec members; and maybe (c) you don’t want them officially associated with debian security coz they’re newbies
<Joey> a) and b) are true
<Joey> However, with security.debian.org being a push mirror, it would be possible to let it be pushed from a second source into a different directory, such as /testing-security/
<aj> hrm
<Joey> However, that would require all architectures to be supported and not just some.
<Joey> all as in “all that are in testing”
<aj> ah, that’s (c) then – they don’t do enough for the official debian stamp
<Joey> ah ok

Assuming those really are the key factors (which might or might not turn out true, but they seem likely) then that’s something that can be worked on. The first point is completely out of my domain – I’ve never been on vendor-sec, and it’s not likely something that ought to be tweaked much anyway; vendor-sec’s a very closed list by design, and for most security work it’s probably a distraction (testing security as is currently operates without concerning itself with vendor-sec, AIUI); the second points purely technical; and the third point is a combination of technical issues and an issue of training, both of which will be hopefully helped a lot by integrating testing-security and security.debian.org. So the easiest point of attack is obviously the second one, since it’s “just” a matter of expanding dak’s featureset. The problem here is embargoed fixes – the security team needs to be able to prepare fixes for vulnerabilities revealed on vendor-sec, but then keep them secret from everyone not on vendor-sec until the publication date. Which is fine, except that dak takes an all or nothing approach to that sort of secrecy: secuirty uploads go into accepted, get built, and then the security team run “amber” to publish the update. But what that means is the accepted directory and its contents can only be visible to people with vendor-sec access, which means only people with vendor-sec access can do security updates. But we do actually have a different queuing mechanism that allows us to keep some things secret, while not hiding everything: thanks to the crypto-in-main transition, the NEW queue acts like that, more or less, only allowing ftpmaster to see the contents of NEW packages until they’ve passed license inspections and the appropriate export notifications have taken place. And actually, we’d already considered extending that sort of behaviour to add an “unapproved” queue that would prevent uploads from entering proposed-updates without the stable release manager’s prior approval. The idea then, would be to have a couple of queues for the security team: an embargoed queue, accessible only to those with vendor-sec access that would store embargoed fixed ‘til they’re ready to be released, and an unembargoed queue for preparation accessible by a wider group, for preparation of fixes that don’t need to be secret. Both queues would need to be autobuilt (and the build logs for embargoed uploads would need to stay restricted, while the build logs for unembargoed uploads would not), which is an added difficulty, since only accepted can be autobuilt presently. But otherwise, it seems basically plausible, both to myself and Joey. Conveniently, Andrew Pollock’s been interested in throwing some money through the AJ Market to see what happens, and since the initial idea we were tossing around fell through due to his move to the US, he’s put up the dosh to make the above happen. Well, strictly the dosh to make me make the above happen, or something. Anyway, I’m already running behind both in implementation and blogging about it, so back to the grindstone… (Okay, perhaps it’s not quite the feature Mark Twain would most like added to dak, but it’s still nifty.)

Yesterday we did meet zufus and gravity for lunch, then mika joined us for bubbles tea. There also was a somewhat unexpected couple joining the group at that point: kiko and his sister. gravity and zufus had to go back to work though, so it was mika, micah, kiko, zuzi and I who went for some Freedom Trail discovery. The weather wasn't that good, and I didn't have too much time either, but we did manage to see a few things around the centre, like the memorial grave yard and a park with tons of fearless squirrels. We headed back to the Acetarium, I quickly took a shower and packed, and we left to the metro, a bit behind on my planned schedule. mika left me two stops before the one I needed to get down in, and instructions to get to the airport from there, using the shuttle. It's not amusing how slow something can appear to be when you're in a hurry. The driver of the bus said it took 5 minutes, but I had to go to the very last stop at terminal E, and it was like an eternity. In the end, I arrived at the terminal at 18:40, when my idea was 18:00, and I ran and ran until I found Air France's desk. There was no queue or anything, only me. During the flight, I tried killing time with War of the Worlds, which was a timekiller but nothing more, and The Island, which I wouldn't recommend unless you care too much about hot female characters. After 6 hours of flight, I was in Paris, apparently on time. But the doors opened much later that when we landed, so my connection started to be endangered. Again, I ran and ran, from one terminal to the other, around the chaos that is CDG. In vain, as my checkin was closed after arriving only 20 minutes before the departure. Much annoyed, I went to the transfer desk, to discover the next flight to Val ncia was cancelled, so the next one was the next morning. The quick alternative via Madrid was full, so I now have tickets for CDG-BCN-VLC, which is extending my travel plan in 7 very nice hours. I have no idea of where my luggage is, but at this point I don't care. I only wonder at what time I'll get sleep. What I do know, is that these fucking idiots claim there is free wireless in this spot of Terminal 2F, but it's not true at all.

It's been a while since I blogged last, but I'll plead lack of time as a defence. Again... :-) Well, after a few weeks in the new job at Amino things are settling down. I'm thoroughly enjoying learning my way around a new codebase, and the people are really good and friendly too. I'm already in charge of the new cvs server at work, which somehow got a Debian installation on it rather than the Fedora that people had been expecting. How could that have happened? On the Debian front itself, I've been a little lax and feeling guilty for that. My advertised debian-cd developments are behind schedule as I haven't had the time to work on it. On the plus side, I've finally diagnosed and fixed the grave bug that was holding the new upstream version of cdrtools out of testing. Once things have migrated through, I'll then be clear to start merging in my JTE code. Then debian-cd upheavals. Yay!

I have read 13 of the books on the list that Clint mentioned. I had begun to write about all the ones which I expected to be there but weren't and then realized I might as well make my own list. So here it is in alphabetical order not in any particular rank. Some entries are actually series.

• 2001: A Space Odyssey, Arthur C. Clarke (But the sequels get progressively worse.)

• Alice In Wonderland/Alice Through The Looking Glass, Lewis Carroll

• Baroque Cycle: Neal Stephenson

• Behold The Man: Michael Moorcock

• Breakfast of Champions, Kurt Vonnegut

• Dictionary Of The Khazars, Milorad Pavic

• Earthsea Trilogy: Ursula Le Guin

• Gloriana or The Unfullfill'd Queen, Michael Moorcock

• Gormenghast Trilogy, Mervyn Peake

• Gravity's Rainbow, Thomas Pynchon

• Jonathan Strange and Mr. Norrell, Susanna Clarke

• Lord Of The Rings, J.R.R. Tolkien

• Naked Lunch, William S. Burroughs

• Ringworld, Larry Niven

• The Computer Connection, Alfred Bester

• The Dune Encyclopedia, Willis E. McNelly (The real backstory to Dune. Brian Herbert should be fed to Shai-Hulud.)

• The Futurological Congress, Stanislaw Lem

• The Island Of The Day Before: Umberto Eco

• The Moon Is A Harsh Mistress, Robert Heinlein

• Ulysses, James Joyce

All pretty damn geeky I think if not stereotypically so.

HP: Upon notification of the screen not behaving properly after the lid is shut and reopened, HP release a BIOS update that fixes it.

Dell: Upon notification of the screen not behaving properly after the lid is shut and reopened, Dell repeatedly fail to return phone calls when expected to, ask for the OS to be reinstalled multiple times, tell me that the motherboard needs replacing and then refer the issue to some internal department with no indication as to how long it will take before anything will be sorted. Total elapsed time so far: over a month.

Thanks, Dell. Thell.

John Carroll:

[T]echnology across software domains should be consistent. There should be a standard, however de facto, that works everywhere. Skills should be portable across those markets, so that someone with experience in desktop software development has a decent chance of developing for handhelds. Everything should just work together, and development across all devices should be relatively straightforward to someone with experience in any one of them. Microsoft has been very good at this. […] I wonder whether the open source world can match that consistency. Granted, there are movements to make things consistent, such as the LAMP set of technologies for Linux, or the Linux Standard Base project. However, the structure of development for open source products militates against the establishment of any consistent standard. Open source development relies mostly on voluntary contributions. People don’t tend to contribute to ALL open source products (that would be impossible). Rather, they concentrate mostly on the products that interest them. This results in a developer version of tunnel vision. Since all that matters is the creation of the handful of products to which the volunteer contributes, what matters is the interests of contributors to a single project, not the interests of technology which should span ALL products.

Great points, though I clearly disagree that the open source community is fundamentally unable to match Microsoft on providing a consistent environment. It’s going to take some work on our part though. One of the reasons Microsoft has historically been so successful is its unique ability to integrate disparate technologies into cohesive wholes, as well as to leave the mechanisms they use to do this just open enough that others can integrate with them to a degree but never so open that they can do so better than Microsoft can. Of course, Microsoft can do this because they develop, own, and control each of the pieces and the interfaces between them. Other platforms (Linux, Java, UNIX, even the Web to a certain extent) are comparative mishmashes for the exact reasons John describes, namely that every project is developed independently, and that there’s comparatively little thought put into how the pieces interact. The only parties that do think about these issues, the Linux distributors, tend to do so after the fact, and, of course, they have no direct control over the projects that produce the pieces to push any integration improvements they might make upstream (in the general case anyway). Furthermore, once multiple solutions in a given problem domain emerge, it’s not always possible to generalize the interfaces when the projects behind them don’t agree with each other on the “right solution”. In other words, because it’s an after-the-fact thing, “not invented here” and general inertia come into play. So, while the technical aspect of the problem is important, namely the interfaces between components (APIs at the source level and ABIs at the binary level), there’s a social aspect here too. We need to encourage the kind of competition at the implementation level that results in technical excellence while fostering a sense of cooperation at the interface level to ensure that, collectively, the various implementations fit seamlessly into a larger, consistent environment. In other words, we don’t want to get in the way of open source project and Linux distro innovation, but we do want to encourage open source projects and Linux distros to innovate in a collaborative fashion so the technology that ultimately wins in the marketplace can be standardized with as few legacy issues as possible (or, in the case where multiple best practices emerge, for legacy reasons or otherwise, we want to collaboratively define a uniform interface, i.e., cooperate on interface, compete on implementation, etc.). That’s the only way we’re going to get there.

So, today has been a day of hardware. Some good, some bad. First, the good. Take one Via EPIA-M motherboard, add a helping of LinuxBIOS and Etherboot and let simmer for a few hours on a low heat. Then add a sprinkling of Phil Hands' d-i preseeding for the final garnish. Tweak to suit your tastes and lo and behold. A neat little motherboard that does proper serial console, boots d-i over the network via tftp and does a completely automated Debian install. Of course, it's not quite that easy. LinuxBIOS requires a few out of tree patches that are still under discussion. The latest version of Etherboot (5.4.1) doesn't seem to work (lots of rubbish on the serial console and no boot), but the version that does (5.2.6) requires a backport of the latest via-rhine driver to avoid leaving the network card in a state that Linux isn't happy with. And Phil's documentation could do with some updating (which I should really email him about). Oh, and add to that the fact that numerous things seem to assume that there is a PC BIOS and that it will support standard interrupts. Yeah, like that's going to happen. Still, I seem to have almost got there and it's quite funky to watch it all happen automatically. Kudos to the d-i team for the ease with which you can preseed. Flush with success I moved on to the next challenge; scanning some forms before posting them off. Nice and easy, normally. If it weren't for the fact that the machine with the SCSI card for the scanner just happens to be the one I stole for the above, and that the new case it's in doesn't allow any PCI cards. Bah. So, find a spare machine, get it into a state where I can actually drive the scanner, scan away. Took far too long. I really need to get it hooked up to my main machine, but there isn't enough desk space to have it close enough for the cables I've got. Perhaps I should crack and buy a USB scanner. I don't really need high quality, just something as an alternative to photocopying. After that trial I've moved on to upgrading my PVR box; I got a new 320GB HD today to replace the 40GB that I'd been using - it was getting to annoying to have to either watch or syphon off various series that I'm recording. While I had the box open I decided to upgrade the motherboard to something that might support wake-on-lan, so this turned into a bigger job than it should have been. :( I'm currently waiting for the data to copy across before trying to stick it back together. Bah. Don't suppose anyone has an S370 P-III better than 667MHz and no better than 1GHz they want to part with cheaply? I'm currently running a Celeron 800MHz (which I think is a slight step up from the previous PIII-667), but a little more oompfh would help with streaming over the network. I'd just upgrade to something modern, but I suspect the extra cost of silent cooling would be a killer.

Weblog Software

I’ve been keeping an online journal over at LiveJournal since 23rd February 2002. To support them I actually bought a “Permanent Account” a while back, an example of putting money behind companies and/or services which are useful - even when it isn’t required. After getting into the swing of things there I started the blog which is used upon Planet Debian and have also been enjoying that. Livejournal is a nice service, and they have great software powering the site. The Livejournal code is available and it is reasonably simple to install. (I’ve setup a few installations of the codebase, and played with them. I admit I’ve never run it in anger). Obviously it is completely overkill to run your own installation of Livejournal - unless, perhaps, you’re a school or a large company. So for my personal blog I use Wordpress. I don’t particularly enjoy using wordpress. To be honest I need only four features/abilities of weblog software:

• The ability to create posts.
• The ability for users to be able to leave comments upon entries.
• The ability to search previous entries.
• The ability to export the blog entries as a feed.

Everything else is utterly overkill. The downside of using Wordpress is threefold:

• Because it is popular there are multiple bots crawling the web and leaving spammy comments.
• Because it uses some funky addressing system any replacement would leave lots of broken links.
• It has historically had a couple of security issues.

I’ve been telling mysefl for the past few months:

• This software has got to go.
• I’ll wait until there is another security hole discovered.

So .. HTML::Template & CGI::Application will at some point come to the rescue. I’m just not sure when. I guess this means I’ve not yet written a bloggin system…. I want to only cover the basics:

• Multiple users.
• Multiple catagories/tags.
• Comments with a sane anti-spam system.

PS - The Sisters of Mercy rock.

If people won't do what you want them to do, changing the law to compel
them to behave in the way you want will not result in them automatically
complying. All it does is grant you the authority to punish them for
behaving in that manner. In many cases, it'll also grant them more
sympathy. If you then proceed to punish them anyway, you run the risk
that more people will start disobeying you as well. It's certainly not a
good way of increasing general happiness.

Yes Branden,
I'm looking at you.

I am a Canadian citizen living in Montr al.
I'm sad and disappointed at the Conservatives party's politicking.  You
appear to be opposing tax cuts.  You appear to be opposing paying down
the national debt (Which was largely created by a conservative prime
minister).
You and your party keeps chanting that the Liberals have lost the "moral
authority" to govern while at the same time acting like children who
won't play along with others.  Nothing in your current behaviour would
inspire me to vote for you.  I would expect Her Majesty's Loyal
Opposition to conduct themselves in a better manner and I feel strongly
that if anyone has failed to govern effectively in this minority
parliament, it's you.
I urge you to stop trying to get an election called two months earlier
than it would come anyway, I urge you to move to a platform of fiscal
conservativeness which supports the sitting government in debt
repayments, and I urge you to begin working on keeping Qu bec in Canada.
I see these as the highest priority items for the House during the last
6 months of this government.
Acknowledgement of receipt is kindly requested.
Tks,
Jeff Bailey

Could anybody tell me how I could get the following behaviour under X11: I want to use an American English keyboard layout because it is more convenient to code. But sometimes, usually in chats, I need my native language characters which are umlauts (latin letters with diaeresis). How could I set up a deadkey combination as a prefix to get those letters? On OS X I could accomplish this by pressing Alt-U (which shows up the two upper dots) and then the desired letter. Something similar would be nice for use under Linux. Could I get such a behaviour just by hacking the keymap or do I need any input engines?