Wednesday, November 29, 2006

FOSS.in 2006 Notes

Notes from some of the talks/discussions, that I attended at FOSS.in, held in Bangalore between 24-26'Nov 2006.

Suparna Bhattacharya, the Linux kernel contributor who works for IBM talked about how kernel hackers are obsessed with making the code simple and the minimalist spirit that pervaded the entire kernel development. She also drew parallels with Buckminster Fuller's idea of ephemeralization - doing progressively more with less. I have noticed that this obsession with internal simplicity and cleanliness is something that is noticeable much more with the better open source projects than in corporate development environments. While it's disconcerting in the beginning, after a while you wonder how development could have been done in any other way. Interestingly Suparna also mentioned that she had used a BBC Micro at school - this tallies with my theory that a lot of good programmers are folks who have had early exposure to computers.

Aaron Seigo, the KDE guy, talked about the planned features for KDE 4 . The Tapioca project aims to make VOIP integration with the linux desktop a reality; SVG support is maturing; The threadweaver project hopes to make writing multi-threaded apps easier; Project Strigi will make desktop search seamless and has a DBUS interface; does meta-data search as well. Aaron also mentioned working with HCI teams to make KDE more usable to regular users - I personally like this idea. The KDE project's UI ideas have always seemed juvenile to me.

Luke Kanies
talked about Puppet - a server automation tool that comes with it's own declarative language. Puppet is written in Ruby and is the result of Luke's frustration with existing tools. 'Developers do not think about manageability', feels Luke. He also added that there are many opportunities to add small pieces of functionality. Puppet has a central server, a portable resource layer and agents on the network that do the actual automation work. The custom configuration language can be versioned and changes tracked easily, since it's plain text. The author of Puppet is very interested in building an open community around this GPL'd software and said that it might take many generations to make the tool great, and he would help create those generations. Resources in Puppet could be anything - Packages, Users, Groups - just about any low level system object. Resources can have multiple low-level providers. The platform specific details are abstracted away. Clients or Agents provide the server with facts about the hosts every thirty minutes or so. Puppet uses SSL everywhere. Though it's been around for only a little while, it is already in production in a couple of places in Europe and US. Stanford Univ. is a paying client. There is no windows version yet and I could not help feeling that the lack of a GUI might be a disincentive to most admins.

This is one area in which I have lots of personal interest, and my current job also involves similar work. I chatted with Luke a bit. I think it's high time we had a good cross platform (*nix, Linux, Mac, Windows)open source tool in this area. It's time to kick the pricey vendors out - they waste our time and money, and have no clue about real requirements. Good luck to Puppet and Luke Kanies. Google videos of his talks elsewhere.

PHP man Rasmus Lerdorf gave a great talk on how to build a good community site and optimizing sites to handle high traffic; even though he used PHP as the example, I thought it was quite general. Rasmus appeared to be a bit upset with the loads of criticism levelled at PHP these days and pointed out the people making negative blog posts about PHP do not realize that Wordpress was a PHP application. I am not sure if this is a good defense of PHP, the language has real problems. But I think it's helpful to remember that for a long time and even now, it's a great way to get things done quickly. As he himself put it, PHP is for weekend warriors. Rasmus actually began his talk by hypothesizing about why geeks are so motivated to work on open source software or even do non-programming community work. He mentioned the case of some users on Yahoo Answers who answer 500-1000 per day questions some times and attributes it to oxytocin, the hormone that gets released during human interaction. Even geeks who rarely see daylight and live in their mom's basement, have the need to connect with other humans, and hence the motivation to take part in net communities. Rasmus feels that it's not what people think about you, but rather what they think about themselves that matter.

Rasmus observed that people should be able to improve a community site by catering to their self interest;he mentioned that no worthwhile community can be built without giving up ownership - if you don't give up control, you will never build a decent community of users. I personally think this works for some categories of software/communities and not for all. Dan Bernstein, the author of Qmail maintains total control over the direction of the project, Napster was Shawn Fanning mainly, and I could go on and these were projects that were so useful that the community ignored the idiosyncrasies of the owner. PHP is now mainly community driven, and became so around 1997; Rasmus' code in PHP is less than ten percent now.

He took apart a small PHP app with the help of Valgrind and showed how profiling an application can be counter intuitive. It was extremely interesting to hear this portion of the talk and I was unaware that valgrind could be put to such use; http_load was used to load test the site. He also mentioned the XML support in PHP5 including Simple XML, Yahoo'd geocoding api, and cautioned against using participation gimmicks like paid blogging. Solve one problem, a clean and intuitive UI were other tips from Ramus. Surprisingly he said that he finds programming to be boring and tedious, and his main interest is in solving problems. Slides here.

Christophe Hellwig, a linux kernel hacker tried to motivate contributions to the linux kernel. Among other things I learnt that one doesn't have to sign over copyrights when submitting code to the Linxu kernel. In contrast contributions to FSF projects need to sign over copyrights to the FSF and do some paper work as well. Christophe said that for result-oriented people there are lots of interesting problems to solve. Lots of modules need help. He explained how corporate contributors to the Kernel were often at conflict with the kernel developers due to pressures from the boss. There was a Sardarji in the audience who asked how many years it generally took to become a kernel developer - the audience laughed for no apparent reason ;-)

Andrew Cowie made an energetic presentation about the structural problems facing open source projects. He outlined at length the problems with centralized version control, build systems, bug tracking systems, and how they all combined to make collaboration between projects difficult. He made the very interesting observation about the committer/non-committer divide created by centralized version control systems. I wonder if DVCSs really solve the problem - you could have your own repository, but what if your changes are never merged into the main repository - C'mon stop pretending that there will be no 'main' repository with distributed version control systems. Besides the problem is how do you discover these repositories - DVCSs do not address the question of discovery at all.

The penultimate day of the conference saw a panel discussion by Sudhar 'Thaths' Chandrashekar; the panelists included Atul Chitnis, Kishore Bhargava, Karunkar Guntapalli, Arun Sharma, Frederick Noronha, KDE developer Sirtaj Singh Kang,and another bloke whose name I didn't get. Atul Chitnis who incidentally does not have too much by way of actual contribution to FOSS projects was predictably the most vocal about young people these days, yada, yada and the death of curiosity. He even made some vague connection between the ability to take apart devices and the ability to hack. In fact there is no such connection. Check out Karl Fogel's excellent article . Only in India do such vague assertions fly without fear of contradiction - it is no better than corporate software environments where the 'get things done' types and 'Excel gymnasts' are regarded as being more useful than good programmers. One recent college graduate finally got fed up with 'oh-young-people-these-days' rants and explained some interesting work they had done at college, that helped reduced the cost of the implementation. He explained that they customised the software extensively and the proprietary choice was very expensive. He also felt that Rs.500/- was a little high for college students. Instead of patiently explaining to a newbie, Atul lit on to this poor chap. One of the panelists was very honest and admitted that he had failed in his MCA exams, unlike lots of pretentious desi FOSS folks who are still enamoured of pedigree. The questions were all predictable, but the answers were better even though the pessimistic tone was a tad too much for me. I can't figure out how Atul arrived at the conclusion that things were going downhill - In fact there have never been so many Indian contributors as now. Atul and co seem to think that being well informed and following every single mailing list out there, and latching on to every passing tech fad is what contribution is all about. Well Atul, with my limited experience in open source projects I can tell you that working on nontrivial patches is real work - the kind of work that requires you to put down your head and do real hard work. Next time you feel like making a claim like this, please do some honest research and go through the commit mailing lists and actual dev lists. I know that you are playing to the gallery and pretending to be the sole saviour of open source in India. Just as life is what happens when you are planning, real open source contribution is what happens when non-programmers who are over the hill are talking about the glorious days of tinkering with modems. Honestly I don't know what the OSS contributions of the panelists other than Sirtaj, and Karunakar are.

It was funny that while one the very visible participants who is a Google employee did not miss an opportunity to let everyone know that he was working with the world's largest organized cult, I heard from another participant at the conference that a very respected open source contributor and O'Reilly author had resigned from Google after spending just three months there. A few days back news of another very respected oss contributor quitting Google trickled in. I know this guy and he is not the one to go around advertising his coolie status even if he works for google. Integrity and self respect count for more than grovelling before the benevolent master.

I attended this event after six years. The first time I met RMS. Nat Friedman's behaviour during his presentation was extraordinarily juvenile - he used the F-word so many times during the presentation that I got irritated. I didn't get a good impression and the many twists and turns of the GNOME project since then have only confirmed the fact that it is run by a bunch of juvenile delinquents. The recent Novell-Microsoft fiasco only serves to confirm my initial impressions.


Best comment/observation of FOSS.in 2006
When a member of the audience, strangely enough a Novell employee, asked if it was not possible to enjoy the freedom without getting into license wars, someone from the back (either Luke Kanies or Andrew Cowie) shouted:
That's a fan club - not an open source software project.

5 comments:

Unknown said...

Very nice write-up; I've pointed to it from my blog.

Anonymous said...

"It's time to kick the pricey vendors out - they waste our time and money, and have no clue about real requirements."

Nice absolutism. Unfortunately, there are a number of vendors out there who do have a clue about real requirements as well as provide actual *gasp* return on investment.

Don't get me wrong. I'd love to see puppet do well; however, I get tired of the use of absolutes and superlatives in relation to the current state of things. That would be like me saying "You've never done your homework". You obviously have looked some...perhaps just not enough.

And yes...getting a systems management platform (or whatever the config-mgmt list wants to call this current product area) working on windows I feel is an imperative. GUIs I don't think are necessary, but I do feel they can be helpful in bringing the tool to the masses / getting buy-in. On the other hand, I think Luke has done a good job so far in getting something out there and iterating.

Unknown said...

how come you are anonymous on my blog?

Rams said...

Kenneth,
I don't have a livejournal login. Don't want to create yet another login and then forget the password.

Anonymous said...

Posted as anonymous b/c i have a policy of not identifying myself on sites that don't have privacy policies. I don't see any privacy policies, etc. so for all i know / knew i'd end up with more junk mail, getting signed up for something, etc.