Ismael
Olea
2004/03/01
You can imagine with how many hapiness I've read this early morning an article by Ian Murdock, Debian founder and Progeny chairman in which he writes about how the Linux distributions should evolve. And I liked all what he said.
Most of my friends here in Spain knows I don't like Debian. Lots of them are even Debian Developers. But this is a personal standpoint and there is no problem further away some friendly discussions: since now I'm not Catholic I don't feel the need of convince the rest of the world with my own ideas ;-)
The interesting part of this «hate» relation is my exploration for trying to find the true weakness and sucesses of Debian. Since I'm not member of their sometimes excesive talibanized community I think I can try to be rigtheus in my conclusions. I hope to find the time in the future to write an article explaining all my conclusions. Always in a constructive way, of course. We are here for helping, not looking for war.
In this moment many of the readers can be asking thenselves why my opinions could be interesting. Well, I have a not so trivial background developing Linux distributions working some time ago for HispaFuentes, an small Spanish Linux company who develop their own version of (now in extinction) Red Hat Linux, which was in their moment a very beautiful product. At the other hand I have done a lot of work in e-publishing with libre documentation using libre software inside internet efforts as TLDP-ES (remember, its name is TLDP-ES but it is spelled L*u*C*A*S). For some years to know I've been thinking about how can be the best and more easy ways to publish documentation, and this process (the publishing cycle) covers all states from authoring to reading. These days I'm trying to put many of the research results published in the web, but I'll need more time.
So, now you know my background. Add to it my technologist interest in trying to understand the world I live in (or more proper said, I work in) and we are ready to show the facts.
Working in the software and documentation publishing cycles, in both cases mainly as an editor, I've found strong similarities between them. If you see it at some distance it could be reasonable since the both are coded knowledge trying to reach somebody interested in use it fully.
Looking with more detail you'll find «suspecting» similarities about technical details in how the «knowledge» should be packaged, formated, tagged with metadata, distributed and used. The most amazing thing is while working in solving documentation problems with the software methods you'll find publishing methods that can improve the way software is managed. I'm amazed by this and I'm looking for support for working full time on this. If you are interested write me ;-)
Other fact is the people need of personalization of the Linux distributions. As I said I worked for a Linux company who created a modified version of Red Hat Linux. The first modifications can be to polish the final quality or the add an extra work on language things, specially translations. There are other points about phisical packaging, presentation of the product and prizing in which I'm not interested here.
The next thing you can find is the convenience of creating and selling distribution profiles, adapted to the final client. This profiling is related i.e. with their corporation image but with the software selection too. The next thing should be to create an automated, or at least mechanized, enviroment for packaging software and create profiles as CDROM ISO images, Internet repositories or, now, CD-live ISO images. Working ahead you can find you need more packages your initial provider don't serve, so you need to add them to your source repositories and need to automate their processing too (three years ago I'd love have the opportunity of using Ximian's Build Buddy :-). In my opinion, what Progeny seems is doing is building a generalized framework which serves in this way.
And here came one of my own discovers: if you are used to work with documentation and are trying to learn how the «physical» and traditional publishing world works, you'll find the previous distribution environment extraordinary similar with how publishing firms works: the create books from sources which can vary: original authors, translators, other publishers, etc. So, don't you think it could be interesting to re-think the way software is published and how is collected? In you are in this business, you should.
Other present trend in Linux distribution are live-cd's. As most know they are great because they are a way to have a working Linux in almost any hardware environment. But is so important how they demonstrates the need of profiles, lots of them, oriented to horizontal or vertical solutions. And I suspect more: maybe is so interesting because there is not an standardised and cheaper way of package and integrate solutions from heterogeneous software sources.
And what about Debian? Some of us are bored of hearing from some of the self-called «debianitas» lots of the arguments about why is the only distribution software of the world the Human Specie should use. As I said at the beginning of this document I plan to discuss the Debian case with more detail but now I want to call the attention in one of the Debian features that most impressed me: the packaging quality process. I was very used to work with the Red Had Linux packages, and I love that distro, but many times I've was disappointed with some quality question you can find unquestionable. For example I've suffered for example the lack of documentation of the RPM .spec syntax, which is outdated for some years. At the other hand I recognise I've never administrated a Debian Linux server and never have an eye to the their packaging sources, but the impression I got, out here, is they have almost perfect skilled software packagers and a collection of software packages ready to use by integrators able to offer solutions to the final client. Debian has a lots of errors and I wish I find the time to describe them in the future, with humor and justice. Few days ago, at OSWC people talk about the Debian Definition: «the perfect way of having up to date an obsoleted linux distribution» ;-)
Now, we can conclude some things.
The publishing cycles for (electronic) documentation and for software are so similars. People working on one of them should investigate in the other and trying to improve their own one with the years (or centuries) of experience of the other.
People at Debian should realize Debian is, today and since some time ago, a publisher of software and they should refine their working cycles having this in mind.
Talking with some Debian developers at OSWC suggested me there are people who starts to understand it. So we can congratulate.
At the Fedora side, I wish they realize too. I love that platform and see a lot of opportunities for them. But if they don't change their mind, offer stable versions and run as a real open project, probably people'll feel the need to change to other platform less interesting but more trustable.
Metadistros (a framework for building live-cd's) and Progeny's componentized Linux platform are integration projects for work with public available software in order of obtein final solutions ready for the client, be final user or a corporate one.
As far as possible, much work should be done here. There is no enough standarization neither speed of integration. Is still complex to get different excellent software pieces for create a solution, even when this could be a relative common profile. Think for example the complexity of setting up a corporate webmail server, or a workgroup one. The time you spend on it is wasted if you can't no reuse the most of it for a second time. That is money lost, higher cost and more time to market.
More work on standarization is still pending. Wish to see good news about this in the near future. At least, we can see positive advances.
PD: By the way. I think Ian is not completely right judging the UserLinux effort. Perens is only follow the Linux standarization effort Red Hat Linux started, which seems to be not well understood and deserves future columns. But as a diference with Murdock's Progeny, I don't compete with Perens' UserLinux, so I can tell it ;-)