Mancoosi Weblog

On October 6th and 7th took place in sunny Lisbon a Mancoosi project meeting. The agenda included talks describing the work already developed in the context of the project by the different partners. There was plenty of time for discussion and to define future research and development directions. The project is beginning to take shape and all our efforts go now to putting into action the results of the meeting.

The next meeting is scheduled for January 2009 and will be held in Nice.

On october 20, 21 and 22 took place in Malaga the Open Source World Conference that attracted quite a large and diverse public, on all aspects of Free and Open Source Software. One of the many tracks was devoted to FOSS projects funded by the European Community, and I gave an overall presentation of the Mancoosi project. It was quite refreshing to participate in an free software event with so many young people around, and the organisation of the conference was absolutely impressive. I think we'll be back next year with some more concrete results to show.

Me and Paulo have just came back from HotSWUp'08, the first ACM workshop on hot topics on software upgrades. There we have presented a paper about co-authored by we two and Roberto. (Remember that, more generally, you can find a good deal of Mancoosi-related papers in the papers section of the main Mancoosi website.)

The (position) paper was about generalities in package upgrades (and their frequent failures!) as experienced in package-based FOSS distributions; interestingly enough, such details have not been properly presented thus far to the interested scientific communities. We believe our work contributed in closing the gap! In addition to the characterization, we point out some problems in state of the art counter measure to upgrade failures. Of course we also hint at the research directions Mancoosi is pursuing to have better upgrade support in future package managers.

The workshop was a pleasant experience and it was very well organized (which is not to be taken for granted in first editions of workshops...); kudos to the organizers for that. Nevertheless, I wasn't expecting to see that there are two big threads in the research community about software upgrades. On one hand there are people working on upgrades in the Mancoosi sense (i.e., how to deploy software on user machines, exploiting the abstraction of packages). On the other there is a huge amount of work on how to upgrade live systems, e.g., how to upgrade the code of a network service, without shutting down the service.

Of course the ideal target is having both safe software deployments and live software upgrades, we simply aren't there yet ...

In the meantime Mancoosi will do its best to address the first part of the issue!

A small Mancoosi-squad (me and Ralf) attended DebConf8, the 9th international conference of Debian project developers.

I delivered a talk about Mancoosi: it is a presentation about EDOS achievements and Mancoosi roadmap, with a particular focus on relationships with the Debian project. Beside that, and as usual for DebConf, several interesting ideas popped up by simply talking with people: edos-builddebchheck has been implemented and work is going on to integrate edos-debcheck with several tools (dput, wanna-build, and Ubuntu's soyouz).

A related news is that we now have a papers section on the Mancoosi website. It contains information about papers and presentations delivered about Mancoosi, and also includes slides of the delivered presentation. For example, you can find there both the slides of the DebConf8 talk, and those of an earlier project presentation delivered in July at the University of Lugano (Switzerland). Enjoy the read!

Our first deliverable, the Mancoosi Project Presentation, is out: you can find it online here, and it contains a detailed presentation of what we plan to do, how we plan to do it, and who we are. For people expert in European administrativia, this is basically a revised version of the Description Of Work of the Mancoosi project.

It was recently brought to my attention that Embedded Debian (Emdebian) is extensively using edos-debcheck for quality assurance purposes.

edos-debcheck is probably one of the most representative utilities of the whole EDOS toolchain: it can be fed with package lists (such as apt's Packages file, in the case of Debian) and asked to check whether some (or even all) packages have unsatisfiable dependencies. If this is the case, something is probably wrong with the distribution represented by the input package list, as it is distributing packages its user will never be able to install.

edos-debcheck is being used in Debian as well, but the use Emdebian is making of it is quite peculiar. Indeed the Emdebian people are using edos-debcheck before an actual upload of bunch of packages to ensure a priori that broken dependencies won't hit the Emdebian archive.

While this would be harder to achieve in Debian (as the uploader as less control on the upload chain, think for example at Packages stuck in NEW: the package universe can change sensibly between the moment they are uploaded and the moment they actually hit the archive), there is room for adding at least some warning at upload time.

Building on the Emdebian scripts, we are working on adding support for such additional checks to dput (the main upload utility used by Debian Developers), probably by the means of pluggable external hook.

From 13 to 15 June, pkgsrcCon 2008 was held at the Technical University of Berlin, and MANCOOSI was there!

We presented our tools, and afterwards there was a good discussion about adapting them for pkgsrc. Look for pkgsrc support in the EDOS and MANCOOSI tools soon!

Being the natural evolution of EDOS, Mancoosi also aims at providing maintenance for the various tools and software developed during its predecessor. In this respect, a lot of work is going on under the hood: stay tuned, we'll keep you posted.

In the meantime, to whet your appetite, we have resurrected the Debian weather service. Looking at it you can get an idea of how the weather will be on a given Debian distribution and architecture. Would it be a good idea to go out today (i.e. perform an install/upgrade attempt) hoping in clear sky (i.e. low uninstallable package count), or would it be better to stay in bed waiting for the storm (i.e. high uninstallable package count) to fade away? Check your Debian weather now to see it!

Daniel points out that Maven is moving to SAT-solver technology too, and this is a second level offspring of EDOS, as it is inspired by the Eclipse P2 project we posted about last week.

So, good ideas travel fast!

One word of warning though for people out there following on the original EDOS ideas:

• we are not necessarily asking that you mention EDOS (that introduced all these ideas back in 2005) all the time (we kind of know that this spoils business advertising a bit); it would be kind to do it once in a while, though :-)
• please get in touch with us at Mancoosi, so we can make sure that the problem description format we are working on for the competition properly encompasses your usage cases, so you will be able to benefit from the competition results.

We have been interacting during EDOS with Daniel Le Berre, when the EDOS WP2 team first pointed out in 2005 that SAT solvers where the right tools to use to check for installability, solving dependencies and conflicts, of software packages in a GNU/Linux distribution.

But software components come in different flavours, and the very same problem arises in a wealth of other fields, so it has been a real pleasure to learn during the Mancoosi meeting in Paris this week that Sat4j (Daniel's SAT solver) will eventually become a part of the P2 project in Eclipse.

This is proof that we are heading in the right direction, as Industry is listening to Academia.

It is also wonderful news for Mancoosi, as we will have lots of potential new participants in the international solver competition which is one of the component of our project.

On May 26th and 27th the first Mancoosi workshop has been held in Paris, with a well packed agenda.

The main goal was to recap the significant EDOS results, review significant related work, and have a few discussions on the work we are doing in the 4 technical workpackages.

We were more than twenty people, and the interaction was useful and intense.

Look forward for the forthcoming deliverables of the project in the next few months!

Welcome to the Mancoosi blog. What is Mancoosi? Let's start with the basics: you pronounce it /mancusi/ in International Phonetic Alphabet.

Then, the meaning: the acronym stands for "MANaging the COmplexity of the Open Source Infrastructure", and the name already says a lot... we are gathering together researchers and free software distribution editors (both from the .deb and .rpm worlds) to produce a next generation set of algorithms to smooth the upgrade process of package based systems. Some of us already provided quite significant contributions in the framework of the EDOS project.

Finally, some administrativia: Mancoosi is a STREP partially funded by the European Community, in the Call 1 of FP7 Objective 1.2. Yes, EC does contribute to improving free software quality :-)

If you want to know more, you can peruse the Mancoosi website, that contains, in web-friendly form, all the information on goals, partners and background.

Take a few seconds to add the RSS feed of this blog to your reader... you will hear more from us soon!