These last weeks, Enrico and me have worked on a GUI frontend for pmnew. At the moment, the prototype is limited to the task of finding new packages to install. The goal of pmnew-gui is to have every little detail designed to be as smart as possible: for example, the list of packages shows at the top the packages that the system thinks you are interested in having comparing your system to thousands other systems. Screenshots http://people.debian.org/~enrico/2007-02/sshot-startup.png
http://people.debian.org/~enrico/2007-02/sshot-audio.png
http://people.debian.org/~enrico/2007-02/sshot-games.png Get it

bzr branch \
http://people.debian.org/~enrico/2007-02/pmnew-gui/

Setup The README gives some details about the setup. In short you will have to do the following steps:

apt-get install python-debian python-deb822 python-xapian \
python-apt python-glade2 python-gtk2
./update-data
./pmnew-gui

It works better if you have a network connection, and if you have popularity-contest installed and running. Feel free to send us feedbacks about the GUI. Links

• "The original brain dump"
• "The commandline draft"
• "Various ideas on improving package managers"

From Dojo Toolkit (a very powerful AJAX toolkit), but should probably go to The Daily WTF...

_getAdjustedDay: function(/*Date*/dateObj)
  //summary: used to adjust date.getDay() values to the new values based on the current first day of the week value
  var days = [0,1,2,3,4,5,6];
  if(this.weekStartsOn>0) 
    for(var i=0;i<this.weekStartsOn;i++) 
      days.unshift(days.pop());
     
   
  return days[dateObj.getDay()]; // Number: 0..6 where 0=Sunday
 

That code is inefficient and stupid on so many levels. For example the if statement... you might be aware that 0 < 0 is false. Yep. I'd prefer something along the lines of

  return (dateObj.getDay() - this.weekStartsOn) % 7;

No arrays were abused during the making of this function. I always thought PHP programmers were the worst, but apparently some JavaScript "coderz" are up to par.

I booked our flights to Austin for South by Southwest on Air Canada last night. We'd originally planned to drive down from Montreal for a nice weeklong roadtrip, then meander along the Gulf Coast out to Miami for our flight to Puerto Rico for the Wikitravel Get-together in mid-April. It would have been a nice trip away from Montreal's gross springtime. But we ended up buying a house, which means we've got to be back in Montreal for April 1st for closing and signing papers and taking possession of the house. Which means we're flying down to Austin and Puerto Rico separately. We're going to take an extra week on the Gulf Coast (Texas). As a last-minute add-on, Maj decided to go to IAsummit 2007 in Las Vegas. So we're swinging through there at the end of March, on our way home. tags: austin sxsw puerto rico gulf coast iasummit house wikitravel get-together travel

Speaking of the house We did our walk-through with the inspector on Monday, which was really great. The current owner, Gilles, was there, and we got to talk a bit. His English is only slightly better than my French, so it was good to switch back and forth. He and his son have lived in the house for years, but he's going to move in with his girlfriend and her three kids, so they need a bigger house. He started the animation studio at Ubisoft nine years ago, so he's got a fun job. The inspector was great -- he really got into the cracks and crevices. And he kept telling us how well the house had been kept up, and what good shape it's in for a house of its age. (Current estimates are that the house is about 100 years old, but nobody knows for sure.) We also put a ladder up on the back terrace so we could go out on the roof, which was just amazing. The neighborhood has mostly small houses, so we have a great view up there. I'm looking forward to building a roof deck this summer. We also just got our financing finalized yesterday. Which means, I guess, that we're really doing this thing. Woohoo! tags: house roof deck gilles ubisoft inspection financing

OpenID bounties What's going on with the OpenID bounties for Open Source implementations that were announced last year at OSCON? As someone who's written one of the main implementations (the OpenID extension for MediaWiki), I'd love to see a bounty paid to the Wikimedia Foundation for that work. The bounty program has some unreasonable requirements, the most onerous of which is that the extension has to ship with the core code. MediaWiki ships without requiring third-party libraries, so the OpenID support has to run as an extension. But I'd still like to see the spirit of the bounty offers followed through, with a bounty given for the MW extension. tags: openid open source mediawiki bounty

After singing the first verse for many years to my kids at night, I chanced upon the full words to this lullaby fairly recently, in an old book on Astronomy for children, Among the Stars by Agnes Giberne, 1885.

::READ MORE

While talking about the ICQ transport on jabber.lusc.de... if you're still using ICQ, please take a look at their Acceptable Use Policy, it contains an important clause: You agree that by posting any material or information anywhere on the ICQ Services and Information you surrender your copyright and any other proprietary right in the posted material or information. You further agree that ICQ Inc. is entitled to use at its own discretion any of the posted material or information in any manner it deems fit, including, but not limited to, publishing the material or distributing it. I don't think it's necessary to comment this.

While talking about the ICQ transport on jabber.lusc.de... if you're still using ICQ, please take a look at their Acceptable Use Policy, it contains an important clause: You agree that by posting any material or information anywhere on the ICQ Services and Information you surrender your copyright and any other proprietary right in the posted material or information. You further agree that ICQ Inc. is entitled to use at its own discretion any of the posted material or information in any manner it deems fit, including, but not limited to, publishing the material or distributing it. I don't think it's necessary to comment this.

Performance of numpy masked arrays This is the code:

import time, numpy, MA
start = time.clock()
a = numpy.ones(1000000)
cur = time.clock()
print "numpy create", cur - start
start = cur
for i in xrange(1000000):
    a[i] = i
cur = time.clock()
print "numpy fill", cur - start
start = cur
a = MA.array(numpy.empty(1000000), mask=1)
cur = time.clock()
print "MA create", cur - start
start = cur
for i in xrange(1000000):
    a[i] = i
cur = time.clock()
print "MA fill", cur - start
start = cur

This is the output:

numpy create 0.01
numpy fill 0.39
MA create 0.15
MA fill 17.06

/me cries

Dell is collecting customer wishes. It's interesting to see that all the top wishes are software related: starting with Linux preinstalled, OpenOffice, no AOL software preinstalled but a plain Vista only, Firefox as default browser or no software installed at all. I'd strongly suggest Dell to consider bundling at least OpenOffice, Firefox, Inkscape, Gaim, Thunderbird, Gimp and some more of the great Opensource software available for Windows (remember: they're free - no extra cost, but some customers will be happy for having a good vector graphics program, or a free office suite...) And I'd definitely like to see Dell offering a "no windows" option; I don't want to pay the Microsoft tax. An OS-less PC is fine for me, but it should be really cheap for Dell to get e.g. Ubuntu preinstalled.

I've decided to split my blog into an English and a German section. This is mostly because I've now also started blogging about some local events that are probably not of much interest to readers from other countries, so I'll blog these in German, but the rest will still be in English. All English posts will still be syndicated on Planet Debian, the German ones will only be readable via my blog or the respective feed. Talking about feeds, I spent some time trying out different feed formats for this blog, since I had some trouble with Planet. The current one (rss2.0) doesn't validate because there is html in the description element but apart from that it seems to be fine, so I'll keep it for the moment.

In my previous post about “Dual boot and full encryption“, I talked about the difficulties of combining full disk encryption for Windows with the option to dual-boot into Linux. Thanks to Jari Eskelinen (who probably is the anonymous user who posted the link to his page in a comment to my original entry), there now is a solution for this problem, at least for DriveCrypt Plus Pack from SecurStar GmbH. I will try this solution soon and it’s likely that I will blog about the results.

Did you ever want to rebuild the archive, but setting up this whole wanna-build/sbuild/buildd stuff somewhat scared you? Then this small script might be for you. All you need is cowbuilder, the rights to build inside a cowbuilder environment and this script. When you call it with a list of packages as argument it will build every package in a clean cowbuilder environment, saving the buildlog (succeeded-$package or failed-$package) and cleaning up the rest afterwards. To speed up things you can run several instances of this script concurrently — lockfiles will prevent the script from building the same package more than once. It is even safe to kill the script and restart it later. It will automatically skip all packages where a logfile (succeded failed)-$package already exists. All you have to do after you killed the script is to remove the lockfile in the LOGDIR and the $package-directory in the BUILDDIR.

#!/bin/sh
# Version: 2007-02-20
LIST=$1
LOGDIR=$(pwd)/"logs"
BUILDDIR=$(pwd)/"build"
BASEPATH="/var/cache/pbuilder/base-testing.cow"
function build  
    PACKAGE=$1
    mkdir -p $BUILDDIR/$PACKAGE
    cd $BUILDDIR/$PACKAGE
    apt-get source $PACKAGE &> /dev/null
    cd $(find . -type d ! -name .)
    pdebuild --pbuilder cowbuilder --use-pdebuild-internal --buildresult .. -- --basepath $BASEPATH &> $LOGDIR/.$PACKAGE
    if [ $? -eq 0 ]; then
        mv $LOGDIR/.$PACKAGE $LOGDIR/succeeded-$PACKAGE
    else
        mv $LOGDIR/.$PACKAGE $LOGDIR/failed-$PACKAGE
    fi
    cd $BUILDDIR
    rm -rf $PACKAGE
 
[ -d $LOGDIR ]   mkdir $LOGDIR
if [ ! -e $LIST ]; then
    echo "No Buildlist given."
    exit 1
fi
while read package; do
    # Create Lockfile (ln -s is atomic and fails if link already exists)
    ln -s $package $LOGDIR/.$package.lock 2> /dev/null   continue
    # Build the package if not already built
    if [ ! -e $LOGDIR/succeeded-$package -a ! -e $LOGDIR/failed-$package ]; then
        echo $(date +%F %T) Building $package...
        build $package
    fi
    # Remove Lockfile
    rm -f $LOGDIR/.$package.lock
done  $LIST

I know my bash scripting skills are shitty, so please send patches after you laughed I’m currently running two instances concurrently for 3 days on my Pentium 2,4GHz HT box and have rebuilt ~70% of the testing archive so far. It would be cool to distribute the rebuilt across a few computers via ssh, so that they’re all building on the same list but using their own cowbuilders. It should be sufficient if all computers shared the same LOGDIR across the network, but I’m not sure how to do this. NFS does not the trick since it cannot guarantee the atomicity of the ln-command.

I've been ill; I had the flu (first time I've ever actually had to take a day off work). And then I spent a week working from Portstewart. Which was a bit odd, as Katherine's living in furnished rented accomodation which means the decor is completely not us (we've always lived in unfurnished places before). Anyway. I went down to London Saturday week ago, and of course there was no direct train from Norwich because Network Rail like to screw over people travelling on weekends. And then I came back from Stansted yesterday and there were no trains from Stansted. So I end up on a replacement bus service to Cambridge (where I'm supposed to pick up a train to Norwich). And the moron driver goes to Audley End (fair enough, we knew it would stop there), reaches J9 of the M11 and decides to go southbound rather than over the junction towards Cambridge. So we end up practically back at Stansted - at J8 he comes off the M11 and rejoins northbound, then drives up to J11 before coming off. I'm not a Cambridge resident, but I'm sure this isn't a sensible route, even factoring in that I don't think you can go northbound at J9. 3 hours of travelling time. I'd have been better on the bus, but I just missed it and the next one wasn't for another 2 hours. And then there's pricing. I'm going to a wedding in Glasgow next month. It was £50 cheaper to fly than take the train (and even given check in times I suspect will be significantly faster). I'm considering a meeting in Manchester at the start of April and flying is cheaper than the train (again by over £50) and in fact given the timings would mean I could actually do it in a day rather than having to sort out accomodation. Flying is even cheaper than driving it seems, and probably faster or at least equal given check in times. I know I'm supposed to be flying less, but it's hard. Give me a viable rail network that's actually competitive on pricing (I'm not even talking about walk on pricing, I'm trying to book in advance with known travel dates/times) and has a decent schedule and I'll use it!

Congratulations are in order to my friend Nicolas Ritoux. He's upgraded Hysterie to use a CMS and CSS. Hysterie is a venerable online magazine here in Quebec, and Niko's converted the site (formerly hand-coded in HTML) to use SPIP, a clever little CMS which seems to be working nicely for him. I think the principal reason he chose it was that it was made in France, however. Anyways, I know he's been working hard on it, since he wanted the site to match pixel-perfect with the hand-coded, table-based, old style. I'm glad he's done; it looks really good. tags: niko spip hysterie cms css

FreeYourID.com I was pretty psyched to see on Scott Kveton's blog the story about the free OpenIDs from the .name registry. Apparently the FreeYourID.com site gives you a free 90-day trial of a .name registered domain, which you can use as an OpenID. I've been using my .name domain as my OpenID for quite some time. I think they're pretty cool. tags: .name opeinid freeyourid.com kveton free

Read receipts I've been having a lot of trouble with email lately. There's been such a flood of stock fraud spam that many of my outgoing emails get caught in other people's spam traps, or lost in the crowd of UCE. I've started requesting read receipts on email that I send to individuals (not mailing lists). It's a simple way for me to make sure that email is getting through to people. I actually like read receipts a lot; I hate writing "I have received your message" messages, and read receipts are a quick and easy alternative. There are some folks who don't like read receipts, and that's fine; they'll just have to put up with me asking through other channels if they got my email. tags: email read receipts spam e-mail tracking

There was an interesting read the other week on LKML where someone or other was trying to say that no kernel config options should be hidden, and any that are should be turned into options that are linked, so that you can "see what you are doing". At the time Linus posted an excellent rant straight back, saying that more options should be hidden, and that the knobs and twiddles that people see should be obvious. I thought at the time that Linus response on this was extremely appropriate, but I didn't expect to be bitten by the opposite approach quite so immediately! I have been trying to build 2.6.20 for the last few days and wondering where in hell has the "Software Suspend" option gone. Well, today I found out that if you disable "Hot Pluggable CPUs" the option will never appear. If Linus has his way this problem would not happen, because the two knobs that people might want to twiddle, which appear superficially unrelated would actually have three settings in truth:

• CONFIG_USER_HOTPLUG_CPU
• CONFIG_HOTPLUG_CPU
• CONFIG_SOFTWARE_SUSPEND

The first setting would be a user-selectable knob, which would ensure that CONFIG_HOTPLUG_CPU was enabled, regardless of the state of software suspend. The last one would ensure that CONFIG_HOTPLUG_CPU and CONFIG_SOFTWARE_SUSPEND were enabled. The current approach is that iff I enable software suspend, then the hotplug CPU option disappears, and OTOH if I disable hotplug CPUs the software suspend option disappears. What I did was to start with a clean, unconfigured 2.6.20 kernel and go through all of the options starting at the top, which is something I tend to do with new hardware. The idea of hot plug CPUs seemed like something I didn't need, so I disabled it. With Linus' approach I would then have enabled Software Suspend a couple of minutes later, and the hidden config option would have been re-enabled behind the scenes. With the way things are at present, I built a good number of kernels and went through quite a bit of hardware blaming before I grepped .Kconfig for SUSPEND and found out what had happened, and which option I needed to enable. It is nice to know that their is such deep sanity at the heart of the kernel development process! Hopefully someone in a position to do so will notice this particular little gotcha and fix it soon, or perhaps someone wanting an "in" to start kernel hacking could go through the kernel configuration and clean up such gotchas. It certainly seems to me that Linus would accept patches for this as being rational.

Vim's Feature of the Day #2: Context-Aware Completion for XML (in 5 minutes!!!) You're editing yet another Debian sub-policy. Of course the policy is written in Docbook XML and (should I even need to mention it?) of course you're using Vim. The Docbook v4.4 specification is huuuuge and you happen to know just a tiny teeny part of it. Hence you're dreaming of XML Spy or Emacs context-aware completion. Shame on you misbeliever!, Vim can do the same!! Here is how to set-up in 9 easy steps context aware completion for Docbook v4.4 with Vim7 in Debian/unstable:

# aptitude install docbook vim-scripts perlsgml
$ mkdir -p ~/.vim/autoload/xml/ ~/.vim/after/ftplugin/xml/
$ cp -r /usr/share/sgml/docbook/dtd/xml/4.4/ /tmp/docbook-dtd/
$ cd /tmp/docbook-dtd/
$ SGML_CATALOG_FILES=/etc/sgml/catalog dtd2vim docbookx.dtd docbk44
$ mv docbookx.vim ~/.vim/autoload/xml/docbk44.vim
$ rm -rf /tmp/docbook-dtd/
$ echo "runtime ftplugin/xml.vim" >> ~/.vim/after/ftplugin/docbk.vim
$ echo "XMLns docbk44" >> ~/.vim/after/ftplugin/docbk.vim

Now just start Vim on an .xml file containing some Docbook markup and let the fun begin:

• in insert mode place the cursor just after a '<' character and type CTRL-x CTRL-o: Vim will present you a choice of opening tags which can be placed there respecting the content model
• in insert mode place the cursor before the '>' character of an opening tag and type CTRL-x CTRL-o: Vim will present you a choice of attributed which can belong to the current opening tag

Warnings: dtd2vim is a recent addition to vim-scripts, you will need >= 7.05. If you start editing a file which Vim is not yet able to recognized as a Docbook file (e.g. an empty file) don't panic, setf docbk is your friend. Enjoy! PS an even higher fun level while editing XML with Vim can be obtained using the "xml" ftplugin shipped by vim-scripts ...

Swedish reaches 80%! A lot of work going on. We're now trying to recover damage done on mysql-dfsg-5.0 and portsentry. I posted a call for updates and translations are arriving at a very fast rate. So, the next target day for the 100% mark for French is February 22nd. Other call for updates are running for a few other packages, for various reasons. As a consequence, I'm flooded with translation updates. I hope I won't mess this up. Today's uploads and NMUs are lirc, norwegian and setserial. Interestingly, this noisy campaign is quietly pushing the idea of a "string freeze" during the freeze before a release. So, it seems that all this energy won't be invested in nothing.

...portsentry and mysql-dfsg-5.0 who both broke debconf translations yesterday. For portsentry this is even more sad as the last upload was done after I nagged the maintainer for l10n updates, which he promised me to fix. He finally did the upload several days after the moment he promised me he would do it....and he just introduced a "fix" to debconf templates in the same upload. :-( As a consequence, I now have one more l10n update to handle and thus I posted a call for updates to translators and I will NMU the package on Wednesday. Same problem for mysql-dfsg-5.0. For this one, translators are used to this as the debconf templates have changed 8 times in a year. NMU planned for February 21st as well. As a consequence, the target is still moving and I really wonder whether we will reach 100% some day. It is now 31 days since I began this campaign.

Ugh. I have some blogging catching-up to do, which I'm going to try to take at a fast clip. I tend to get bogged down in the daily details, which can make getting my info down electronically kind of difficult.

Trip to New York City Maj, Amita June and I drove down through upstate New York for LinuxWorld OpenSolutions Summit. It was a good drive -- really pretty going through the Adirondacks in winter. There were icicles hanging dozens of meters from stone cliffs -- really dramatic. We spent Tuesday night with our friends Jen Bohner and Massoud Marzban. Jen is a friend of mine from high school, and Massoud is her husband. They live in Sunnyside, Queens now, after a year stint here in Montreal just before Amita was born. We miss having them around here, but they seem to have settled in well in NYC. Jen is working way too hard on a master's of education while teaching 5 classes during the day in New York public schools. Despite this, they found time to make us a great dinner, open a bottle of wine, and spend the evening chatting. Really decent people. tags: jen massoud drive montreal travel queens

LWOSS The next morning, Valentine's Day, I got up early to go into Manhattan for LWOSS. I called for a car service that Jen and Massoud use, but they failed to show up, and when I called back they hung up on me. Deborah's Car Service, there. Make sure to miss them if you're in the area. Anyways, I managed to flag down a cab and get into Manhattan anyways. It wasn't too much fun -- there'd been a big snowstorm the night before, and traffic was gnarled and bitter. Still, I got to the event hotel -- the New York Marriott Marquis on Times Square, got checked in, and registered for the event just fine. The keynote speaker was security expert Bruce Schneier, who gave an extremely insightful talk about economic laws as they apply to computer security. It was very interesting, if hard to follow at times -- it's not always easy to keep economic theories and security practices juggling in your brain at the same time. I met up with Don Marti, who'd invited me to participate at LWOSS, in the speakers' lounge. He had assembled a great team of interesting people who know a lot about Linux and Open Source: John Mark Walker, Jeremy Allison and Jerry Carter of the Samba project, and Jeremy Garcia of LinuxQuestions. Sadly, the snowstorm had canceled the flights of two more participants, Donald Becker and Sascha Chua. (Those weren't the only two people who couldn't come because of the flying problems. I overheard dozens of conversations on cell phones of people whose colleagues couldn't make it for the show. In retrospect, it's pretty surprising that we had the event at all. Schools and businesses were closed all across New York on the 14th and 15th.) Don's plan was simple: to hold an "unconference" at the heart of the conference. We took a section of the showroom floor, set up some chairs and a big monitor for demoing, plus a whiteboard, and just started some interesting conversations. I talked about Web APIs, since I wrote an article on Web APIs for LinuxWorld.com in December. Other people talked about their own projects, about Linux distros, about tools in the Open Source toolkit, about the process itself, and all kinds of things. It was really fun and productive. (I'm realizing now that I should put Don in touch with Ted Ernst, who's been instrumental in organizing three of the Open Space events I went to in the last year. It would be really interesting to have an Open Space event the day before or after the main conference... making mental note, here.) Otherwise, the event was really pretty good. I talked a lot with Jeremy Garcia, who started using MediaWiki about the same time we started using it for Wikitravel. LQ is a really great project, and a nice mix of different community platforms. I managed to luck into an LQ t-shirt, which I'll wear proudly. The other big news, for me, was getting a loaner Nokia N800 to use during the event. The N800 is an Internet tablet, about the size of a cell phone, but without phone features. I liked it so much that I asked the Nokia team to loan me the device for a long-term review, and they agreed. So I took it home with me. Meanwhile, Maj and Amita June had a nice time tooling around New York. They went to the American Museum of Natural History on Thursday, which was pretty great -- Amita loved looking at the animals. We had dinner on Wednesday at a pub off of Times Square called The Perfect Pint -- although the food was a little more upscale than pubby. Thursday night we got a babysitter through the hotel and had drinks at the Broadway Bar (in the hotel) and dinner at The View (also in the hotel). The View is a rotating restaurant at the top of the Marriott Marquis -- pretty fun to do. tags: lwoss don marti unconference manhattan marriott marquis linuxquestions

UbuCon On Friday, I went to the Google HQ in Manhattan for TheUbucon, another (!) unconference, this time about Ubuntu and the people who love it. I got to hear Mako talk about the Ubuntu community process, how it works, and how to get involved. It was pretty impressive. I also did some key-signings and talked to a lot of interesting people. Including Kevin Mark, who had lots of comments on my blog. Hi, Kevin! I had to take off just as the event was getting started, to go pick up Maj and Amita June at the hotel and check out. We took a cab back to Queens, where we'd stashed the car so we didn't have pay outrageous Manhattan parking fees. Sadly, someone had taken the opportunity to key our paint job. I love you back, New York. tags: ubucon google kevin mark

Drive home I wanted to drive back through Connecticut, Massachusetts and Vermont on the way back, for variety, so we turned east and headed up through Westchester County. It turned out to be a real mistake -- traffic on a Friday afternoon out of New York is apparently pretty miserable. We took 4 hours to get to Hartford, by which time the baby was miserable and the rest of us were feeling pretty punked out, too. We ended up in the Motel 6 in Enfield (Connecticut) where we ordered pizza, took showers, and just got our heads back together. Today we got up early and had a nice drive through Western Mass and Vermont. The little northern New England towns were pretty beautiful in the snow, and we passed a lot of barns and farmhouses that looked like movie sets. We stopped in Burlington for lunch, and had a leisurely drive over the border and back home. When we got to our house, I had to shovel drifted snow -- about 3-4 feet deep -- out of our parking space so I can pull the car in. But otherwise the house is in good order. I'm glad to be back. tags: connecticut massachusetts vermont burlington snow home