I haven't posted one of these in a while, but now that I'm finally caught up with writing reviews (although not posting them yet), I've been feeling about reading again. And of course the backlog always grows. Let's see if I can figure out what I've picked up since the last post. Catherine Asaro The Phoenix Code (sff)
Annie Bellet Justice Calling (sff)
Annie Bellet Murder of Crows (sff)
Annie Bellet Pack of Lies (sff)
Zen Cho Sorcerer to the Crown (sff)
J.J. Gribble Steel Victory (sff)
Judith L. Herman Trauma and Recovery (non-fiction)
Janis Ian & Mike Resnick Stars: The Anthology (sff anthology)
Megg Jensen Hidden (sff)
T. Kingfisher Bryony and Roses (sff)
T. Kingfisher The Seventh Bride (sff)
Nancy Kress Crossfire (sff)
Edmond Lau The Effective Engineer (non-fiction)
Yoon Ha Lee The Fox's Tower and Other Tales (sff collection)
Lozzi Roma Rome and the Vatican (non-fiction)
Elizabeth Lynn The Dancers of Arun (sff)
Elizabeth Lynn The Northern Girl (sff)
Rhonda Mason Empress Game (sff)
Vonda N. McIntyre Starfarers (sff)
Linda Nagata Memory (sff)
Jody Lynn Nye Strong Arm Tactics (sff)
Cat Rambo Near + Far (sff)
Kristine Kathryn Rusch The Diving Bundle (sff collection)
Kristine Kathryn Rusch Recovering Apollo 8 (sff)
L. Shelby Serendipity's Tide (sff)
L. Shelby Treachery's Harbor (sff)
L. Shelby Fealty's Shore (sff)
K.B. Spangler Digital Divide (sff)
Judith Tarr Forgotten Suns (sff)
Greg van Eekhout California Bones (sff) Hm, that's quite a lot. Most of it is various ebooks I picked up from various places, including a Humble Bundle. I've been impulse-buying a lot of stuff based on James Nicoll's reviews. The travel guide was a gift from a friend from a vacation in Italy.

Last weekend, I (knok), Hideki (henrich) and Yutaka (gniibe) met with John Paul Adrian Glaubitz (glaubitz). In the past, I had met with another Germany developer Jens Schmalzing (jensen) in Japan. He was a good guy, but unfortunately he gone in 2005. I had an old OpenPGP key with his sign. It is a record of his activity, but the key is weak nowaday (1024D), so I stop to use the key but don t issue revoke. Anyway glaubitz is also a good guy, and he loves old videogame console. gniibe gave him five DreamCast consoles. I bring him to SUPER POTATO, a old videogame shop. He bought some software for Virtual Boy. DebConf 2015 will hold in Germany, I want to go for it if I can.

The last few days I have been experimenting with the batman-adv mesh technology. I want to gain some experience to see if it will fit the Freedombox project, and together with my neighbors try to build a mesh network around the park where I live. Batman-adv is a layer 2 mesh system ("ethernet" in other words), where the mesh network appear as if all the mesh clients are connected to the same switch. My hardware of choice was the Linksys WRT54GL routers I had lying around, but I've been unable to get them working with batman-adv. So instead, I started playing with a Raspberry Pi, and tried to get it working as a mesh node. My idea is to use it to create a mesh node which function as a switch port, where everything connected to the Raspberry Pi ethernet plug is connected (bridged) to the mesh network. This allow me to hook a wifi base station like the Linksys WRT54GL to the mesh by plugging it into a Raspberry Pi, and allow non-mesh clients to hook up to the mesh. This in turn is useful for Android phones using the Serval Project voip client, allowing every one around the playground to phone and message each other for free. The reason is that Android phones do not see ad-hoc wifi networks (they are filtered away from the GUI view), and can not join the mesh without being rooted. But if they are connected using a normal wifi base station, they can talk to every client on the local network. To get this working, I've created a debian package meshfx-node and a script build-rpi-mesh-node to create the Raspberry Pi boot image. I'm using Debian Jessie (and not Raspbian), to get more control over the packages available. Unfortunately a huge binary blob need to be inserted into the boot image to get it booting, but I'll ignore that for now. Also, as Debian lack support for the CPU features available in the Raspberry Pi, the system do not use the hardware floating point unit. I hope the routing performance isn't affected by the lack of hardware FPU support. To create an image, run the following with a sudo enabled user after inserting the target SD card into the build machine:

% wget -O build-rpi-mesh-node \
    https://raw.github.com/petterreinholdtsen/meshfx-node/master/build-rpi-mesh-node
% sudo bash -x ./build-rpi-mesh-node > build.log 2>&1
% dd if=/root/rpi/rpi_basic_jessie_$(date +%Y%m%d).img of=/dev/mmcblk0 bs=1M
%

Booting with the resulting SD card on a Raspberry PI with a USB wifi card inserted should give you a mesh node. At least it does for me with a the wifi card I am using. The default mesh settings are the ones used by the Oslo mesh project at Hackeriet, as I mentioned in an earlier blog post about this mesh testing. The mesh node was not horribly expensive either. I bought everything over the counter in shops nearby. If I had ordered online from the lowest bidder, the price should be significantly lower:

Supplier	Model	NOK
Teknikkmagasinet	Raspberry Pi model B	349.90
Teknikkmagasinet	Raspberry Pi type B case	99.90
Lefdal	Jensen Air:Link 25150	295.-
Clas Ohlson	Kingston 16 GB SD card	199.-
Total cost		943.80

Now my mesh network at home consist of one laptop in the basement connected to my production network, one Raspberry Pi node on the 1th floor that can be seen by my neighbor across the park, and one play-node I use to develop the image building script. And some times I hook up my work horse laptop to the mesh to test it. I look forward to figuring out what kind of latency the batman-adv setup will give, and how much packet loss we will experience around the park. :)

It s been a while now since the 2008 Summer of Code ended. This year, twelve (?) projects were selected. That s twelve students working full time on a Debian-related project during the summer. The Google Summer of Code has sometimes been criticized in the past for having a poor student-developer retention rate inside the host projects. One of the goals of the program has always been to bring new people to budding or established free software organizations and it s a pity that some would leave the project as soon as the program ends. On the other end, poor integration of created code within the project leads to work that is hard to merge in, or worse, doesn t get merged in at all. That s a waste of time and resources and a probably cause of global warming as well. Hopefully, it s not always the case. Some people choose to stay committed within the organization in the long-term. Useful code gets merged in and pushed to the public. I am going to give a talk about this at FOSDEM (go to FOSDEM!) so I m giving you a little preview. I need your help to collect information for my talk. As you know, information is always hard to come by with these kinds of projects so anything can be useful. Without further ado, let s have a look at the cast of the Debian Google Summer of Code 2008: Netconf, a network configuration management system Presentation Netconf is a network configuration management system designed with modern network infrastructures and the needs of roaming users in mind. It is a personal project of Martin Krafft that he started in 2007. He did some presentations about it that you can find on the dev website. The project proposal was introduced by the mentor. The work was mainly about completing the roadmap items for version 1.0. Most of the design was done and code fleshed out. The stated goal was to have netconf ready for lenny. Martin noted that due to lack of regular free time, he couldn t reach that goal by himself. Student Jonathan Roes was a computer science graduate student from the University of North Carolina at Charlotte. He had programming experience as a hobby for a long time and wrote some free games and libraries for the Nintendo DS and some proprietary webapps. He submitted a few little patches in mid-march right after the publication of accepted mentoring organizations and went on to work from mid-may to mid-august. He wrote a lot of code right into the trunk since the whole project was a prototype. Result The last commit by Jonathan was also the last to date in the main netconf git repository. No further progress has been made and obviously netconf didn t get into lenny. I couldn t find any further public involvement of Jonathan within Debian. The ultimate Debian database, all things Debian in a SQL database Presentation The Ultimate Debian database wants to reunite all Debian data sources in a SQL database The project was mentored by Lucas Nussbaum and co-mentored by Stefano Zacchiroli and Marc HE Brockschmidt. The project proposal was introduced by Lucas. Student Christian Neronus Von Essen is.. well, there wasn t much information readily available on him. Result The whole project is coded up and working well with a whole bunch of data sources. There will be a talk at FOSDEM about this so I ll leave it to Lucas to talk about it in detail. I couldn t find any further public involvement of Christian within Debian. Security-beta, a beta testing for Debian security updates Presentation The task is to improve the quality assurance process for security updates by providing a public security update beta test program in addition to the existing QA done for security updates. During the preparation of security updates, there s an inherent delay between the initial upload of the fixed packages and the time until the packages have been built on porter machines. This time gap will be used for a new security update beta program. The project was supposed to be mentored by Moritz M hlenhoff. Student The project was supposed to be done by Nico Nion Golde. He is studying computer sciences at the Technical University of Berlin. He s also a DD. Result There s no nothing. Nico, what happened ? And obviously, he s still developing for Debian. Debgraph, a generic infrastructure for the development of packages management tools Presentation In a large software ecosystem such as Debian Linux, there is the potential for dependencies among software packages to create complex management and technical problems. For example, dependency loops (cycles) in which a package directly or indirectly depends on itself can confuse package management tools as they determine the proper order of package installation. debgraph helps developers to solve this problem by enabling generic queries (e.g., Give me all the nodes that depend on package X ) against the graph of packages and thus automating much of the manual labor that is typically involved in resolving dependency problems. The project was mentored by Robert Lemmen, who introduced the project proposal. The project was already started and the C++ code foundation was done by the time it was proposed. Student The project was executed by Adam Jensen, research assistant in the Software Engineering and Network Systems Laboratory at Michigan State University. Result Adam maintained a blog about his progress and finished ahead of schedule. However, the resulting work seems to be unused, which is a pity since the code could be used within other programs (package managers?). I couldn t find any further public involvement of Adam within Debian.
That s all for now. The information is quite fragmented I admit. Most of it was pulled from Google, mailing lists, commit logs, blogs, whatever. If some projects are lacking in information here, it s because I couldn t find it readily (which is an issue in itself!). In my next post, I ll try to analyze the success and failures to extract some insight. Teaser: pet projects!
If you re a student or a mentor mentioned above, feel free to fill any of the blanks in my report. It s much appreciated. You re not a student or mentor mentioned above and have an opinion on how to improve the next Debian Summer of Code ? Feel free to comment.
Sledge, ping! Here s a list of projects to be described in my next posts:

• Debian NAS, improve support of Debian on NAS devices
• Cran2deb, generate Debian packages from R packages
• Mergemaster, interactively merge changes in configuration files
• PAS NSS Debian Installer, improve support of PAM and NSS at install-time
• Jigdo-ivory, a JavaScript Jigdo client
• Aptitude-gtk, usability and GTK+ GUI for the Aptitude package manager
• Lintian for fuller automated setups
• Debexpo, a generic web-based package repository

Currently there is a debate about censoring the Internet in Australia. Although debate might not be the correct word for a dispute where one party provides no facts and refuses to talk to any experts (Senator Conroy persistently refuses all requests to talk to anyone who knows anything about the technology or to have his office address any such questions). The failures of the technology are obvious to anyone who has worked with computers, here is an article in the Sydney Morning Herald about it [1] (one of many similar articles in the MSM). I don’t plan to mention the technological failures again because I believe that the only people who read my blog and don’t understand the technology are a small number of my relatives - I gave up on teaching my parents about IP protocols a long time ago. One of the fundamental problems with the current censorship idea is that they don’t seem to have decided what they want to filter and who they want to filter it from. The actions taken to stop pedophiles from exchanging files are quite different from what would be taken to stop children accidentally accessing porn on the net. I get the impression that they just want censorship and will say whatever they think will impress people. I have previously written about the safety issues related to mobile phones [2]. In that document I raised the issue of teenagers making their own porn (including videos of sexual assault). About four months after writing it a DVD movie was produced showing a gang of teenagers sexually assaulting a girl (they sold copies at their school). It seems that the incidence of teenagers making porn using mobile phones is only going to increase, while no-one has any plans to address the problem. The blog www.somebodythinkofthechildren.com has some interesting information on this issue. Two final reasons for opposing net censorship have been provided by the Sydney Anglicans [3]. They are:

1. Given anti-vilification laws, could religious content be deemed illegal and be filtered out? Could Sydneyanglicans.net be blocked as illegal if it carries material deemed at some point now or in the future as vilifying other religions? If it s illegal in Vic say, and there isn t state-based filtering (there wont be), will the govt be inclined to ban it nation wide?


2. Given anti-discrimination laws, if Sydneyanglicans.net runs an article with the orthodox line on homosexuality, will that be deemed illegal, and the site blocked? You can imagine it wouldn t be too hard for someone to lobby Labor via the Greens, for instance.

So the Sydney Anglicans seem afraid that their religious rights to discriminate against others (seriously - religious organisations do have such rights) will be under threat if filtering is imposed. I was a bit surprised when I saw this article, the Anglican church in Melbourne seems reasonably liberal and I had expected the Anglican church in the rest of Australia to be similar. But according to this article Peter Jensen (Sydney’s Anglican Archbishop) regards himself as one of the “true keepers of the authority of the Bible” [4]. It seems that the Anglican church is splitting over the issues related to the treatment of homosexuals and women (Peter believes that women should not be appointed to leadership positions in the church to avoid “disenfranchising” men who can’t accept them [5]). It will be interesting to see the fundamentalist Christians who want to protect their current legal rights to vilify other religions and discriminate against people on the basis of gender and sexual preference fighting the other fundamentalist Christians who want to prevent anyone from seeing porn. But not as interesting as it will be if the Anglican church finally splits and then has a fight over who owns the cathedrals. ;) A comment on my previous post about the national cost of slow net access suggests that Germany (where my blog is now hosted) has better protections for individual freedom than most countries [6]. If you want unrestricted net access then it is worth considering the options for running a VPN to another country (I have previously written a brief description of how to set up a basic OpenVPN link [7]).

• [1] http://www.smh.com.au/articles/2008/11/11/1226318639085.html?page=fullpage


• [2] http://doc.coker.com.au/safety/mobile-phone-safety/


• [3] http://www.sydneyanglicans.net/forums/viewthread/3673/


• [4] http://www.smh.com.au/articles/2008/06/22/1214073053814.html


• [5] http://www.abc.net.au/worldtoday/content/2004/s1213397.htm


• [6] http://etbe.coker.com.au/2008/11/17/national-cost-slow-net-access/#comment-16729


• [7] http://etbe.coker.com.au/2008/05/24/ipsec-is-pain/

Yesterday, 10 July, was the due date for the second milestone of my work on DebGraph. I am happy to report that we are roughly two weeks ahead of schedule, so meeting this milestone was not a cause for worry. We now have support for the following graph operators:

• Difference
• Intersection
• Filter (by properties)
• Find Cycles (via Tarjan's algorithm)
• Find Dependencies
• Find Reverse Dependencies
• Symmetric Difference (XOR)
• Union

The next milestone includes the development of a high-level language (or integration with an existing extension language) that streamlines the construction of complex queries using the operators listed above. We can build arbitrarily complex queries using the C++ operators, but dealing with the static typing and compiler toolchain can be very clunky. As such, I have spent the past week working on Lua bindings for DebGraph, which will enable us to query the graph of Debian packages from a clean, dynamically typed language. Lua has a powerful C API that exposes the Lua stack and lets us move DebGraph information to and from the Lua interpreter. I'm writing documentation that outlines how to use DebGraph from both C++ and Lua in order to make this work accessible to more folks. Sneak peak
As an example, we can utilize the FindCycles operator in Lua as follows:

libdg = package.loadlib("./libdebgraph.so",
"luaopen_libdebgraph")
libdg()
LoadPackages('cache')
-- 'g' is the main graph of unstable binary-arm packages
cycles = FindCycles(g)
print("Found " .. #fc .. " cycles")
for comp_key,comp in pairs(cycles) do
    comp_nodes = ""
    for node_key,node in pairs(comp) do
        prop = GetProperty(node, "Package")
        comp_nodes = comp_nodes .. prop .. " "
    end
    print("* " .. comp_nodes)
end

The above Lua script produces the following result:

reading cache/unstable/main/binary-arm/Packages
Found 61 cycles
* debconf-english debconf-i18n debconf
* perl-modules perl 
* cdebconf fontconfig-config fontconfig gsfonts-x11
libcairo2 libfontconfig1 libgtk2.0-0 libpango1.0-0
libpango1.0-common libxft2 ucf x11-utils xutils
[...]

These are the package names in the strongly connected components, shown in alphabetical order. There is still a lot of work to do before the Lua binding will achieve feature parity with the core library, but we have now laid the foundation (and a brick or two).

The overall situation (admittedly a bit of a stretch):

(The soil depends on the chicken to keep it fertilized. Work with me here.) There are an infinitude of cycles that could be extracted from this, but we're only interested in the set of shortest-length cycles. Exhibit A:
Exhibit B:
And the classic philosophical gem, Exhibit C:
Is the problem solved, then? Not quite. The current DebGraph cycle detection operator works well for dependencies, but we also want to discover cycles of conflicting packages. Furthermore, one of our documented use cases involves finding nodes that depend on and conflict with one another.

The GNU C library (packaged as libc6 in Debian) is the bedrock of a lot of code. What are the packages that it depends on? (Yo, DebGraph!) Original image [997 kB] libc6 is one of the well-connected vertices in the lower-left corner (third quadrant, if you like). For graphs like this one, we would like DebGraph to identify dependency cycles (loops). For example, libc6 depends on libgcc1, which in turn depends on libc6. This can be seen in the graph above by applying some careful scrutiny, but why not let the computer work its magic for us? (To be continued.) As a bonus, who depends on arping 2.05-2 for i386?

Watched De Gr nne slagtere (great black comedy even if it moves a bit slowly. I think if you have to chose you'd better watch Adams bler with which it shares writer/director Anders Thomas Jensen and a big part of the cast). And this just has to be said: I can't believe anybody could call eXistenZ “quite a good see”. One of the worst movies I've ever seen, on a level with “Tweed” (a late 90s Bond parody I distinctly remember having seen but can't find on imdb right now.) I agree with Adeodato's assessment of Billy Elliot, though.

Another week has been spent in a repetition course in the military. Readers of my blog will know that this means playing poker (I lost about CHF 7 this week — we just play for fun at CHF 5 for 50 chips) and watching movies. Tron and Kill Bill Vol. 1 from my collection, and the extremely funny Adams bler which I didn't know before. The story centers around a neo-nazi serving some time in some kind of rehab in a church where, it is soon discovered, the priest in charge of the whole thing is not that much in charge, even of himself. I'm told that I also should watch I Kina spiser de hunde by the Danish Anders Thomas Jensen (writer and director of Adams bler, writer of In China they eat dogs.) Also recently watched Casino Royale (the recent Bond movie. I haven't seen the 1967 one yet.) and The Gold Rush: the 1942 version with added spoken narrative and music by Chaplin. The DVD does contain the original 1925 silent movie, too, where especially the ending is a bit different. And finally lots of episodes of , ! (Nu, pogodi!), which was the Soviet equivalent to Tom and Jerry (not a clone, it is entirely different.) I can highly recommend this to everybody who likes cartoons.

I decided to move the old 1-RU Pentium III server (the thing providing all the storage for my MythTV box via ATAoE) from under the bed in the spare room to the linen cupboard. It had a brief stop in the wardrobe in the spare room, but it didn't really help with the noise. Sarah was a concerned about the heat in the linen cupboard, as this is where the patch panel is, so caesar is already in here, along with an Ethernet switch, an ADSL modem, a wireless access point, and the Vonage ATA. There's a lot of DC plug packs generating heat, so it was a reasonable concern. I figured in this day and age, someone must make some sort of USB temperature sensor, and after some searching around, I discovered the DLP-TEMP-G, which seemed to be about the right price, doing what I wanted. I'm going to say the web page was a bit ambiguous, but it was probably just late at night, because I read the bottom as having the option to buy it from Mouser, or via PayPal directly with DLP Design. So I went down the latter path, because I generally prefer to deal directly with the manufacturer. Turns out what I was actually ordering was just a "Test Application", so that email I'd received the day after I placed my order for two units was all I was going to get. I figured this out after a week or so of wondering when it was going to arrive. They were really good about it, and refunded me my $40, even though I'd received said software, and I placed a new order with Mouser for what I really wanted. Now this time, I just didn't read things thoroughly enough, although I'll still say that things were a little ambiguously worded. The page I've linked to above does say "DLP-TEMP-G and 1 DS18B20 sensor $25.00", but when reading datasheet, it goes on about three sensors, and coming with one that isn't soldered on so that you can optionally run a cable between it and the board. I somehow interpreted this as meaning there were two sensors on the board, and one loose. Not the case. The board supports having up to three DS18B20 sensors attached to it, however you please. It comes with one unattached. In hindsight, why would you want two temperature sensors directly on the board? So the moral of this story is I just can't read. Anyway, I got home from Santa Monica last night, and the goods had finally turned up, so I had a play. The DS18B20 just looks like a transistor. I initially ignored it, and just shoved the USB board in caesar, and it happily recognised it

usb 1-1: new full speed USB device using uhci_hcd and address 8
usb 1-1: configuration #1 chosen from 1 choice
drivers/usb/serial/usb-serial.c: USB Serial support registered for FTDI USB Serial Device
ftdi_sio 1-1:1.0: FTDI USB Serial Device converter detected
drivers/usb/serial/ftdi_sio.c: Detected FT232BM
usb 1-1: FTDI USB Serial Device converter now attached to ttyUSB0
usbcore: registered new driver ftdi_sio
drivers/usb/serial/ftdi_sio.c: v1.4.3:USB FTDI Serial Converters Driver

I then fooled around with minicom, and discovered that the little transistor thing I'd been ignoring was indeed the temperature sensor, as I got a reading of zero back (when using this program I found on the 'net). So I went to bed, and this morning did a bit of messing around with the sensor, and with a bit of creative bending, I've got it sitting in the S1 holes without requiring any soldering. It tells me the linen cupboard is about 44 degrees Celsius. Warm, but I don't think it's in any immediate danger of bursting into flames. Wouldn't surprise me if some of the gear in there isn't too keen about the temperature though. At least we won't have to worry about mold. Next step is to convince cacti to graph it, and nagios to monitor it, and we're in business. Here's a little Python program I knocked up to grab the temperature. pyserial is nice. Read on, Macduff!

I just submitted my talk proposal for YAPC::EU, which will be held in Vienna in late August. The topic? Integrating Perl in a wider distribution: The Debian pkg-perl group.
I took part in YAPC::NA (Yet Another Perl Conference - North America) in 2001, in Montreal, and YAPC::EU (you guessed right: Europe) in 2002, in Munich. While in Munich, I met Debian's Erich, Weasel and the late Jens. It is a very nice conference with all kinds of Perl-heads, apt for different experience levels. the talk I am proposing will be about what do we work for in Debian, how can we get a better synergy from our upstream group and (oh, this point is quite itchy - specially in strongly opinioned communities such as Ruby's! Perl people are quite nice to play with, however, but still...) what do we (as Debian maintainers) request from them in order for life to be smoother.
Of course, I only submitted the talk. The CFP has just been launched, and so far, mine is the fourth talk offered - It can still be rejected (as it is not really related to Perl development, the heart of YAPC), but I guess it will be deemed interesting by the Perl monks reviewing them.
In any case, hope to see you in Vienna as well!

I have the impression that OpenOffice is still busy with cleaning up the code inherited (and e.g. porting to GTK2 and such things). Given the screenshots posted in the much respected Office UI blog by Jensen Harris, I think OpenOffice will run into trouble when Microsoft actually releases their new Office. Microsoft seems to have a really neat and productive UI there - there is only one thing really wrong with it: it's completely different than anything else people are used to (and people have learned to use). While there will obviously be migration issues, companies will probably delay switching to the new office to save licensing costs and such things, marketing people will be all crazy about the new features in there. You might remember the research with the aggressive title "PowerPoint makes you stupid"... actually this will become even more true with the new Office, I fear. It comes with default styles for common diagrams used in presentations, like a flow sequence or a circle sequence. While these aren't particularly smart things, they'll be presented even more prominently (because the gfx is so great), and the actual contents will probably be pushed back even further... Anyway, these new styles look really hip, so people will be eager to use them. And use them. And use them again. And use them really big (eventually forgetting about the contents behind them)... and OpenOffice can't keep up with that so far. Maybe OpenOffice needs a huge scale UI contest. Maybe with some abstraction scheme to be able to run different UIs on the same engine. OpenOffice in particular should try to become a technology leader, and not just try to copy Microsoft as much as possible... the UI changes in Impress are said to be pretty similar to what people expected when they think PowerPoint. There are good reasons to stick with the established UI standards (even when that might mean 'cloning' the UI) since that keeps migration costs low. But I think it's a pity OpenOffice can't show many large technological improvements to put more pressure on Microsoft. Ah, and yes, I guess many people at OOo are well aware of that, and like so many projects (commercial ones probably even more than FLOSS) they're just short on development resources...