The first two days of this conference were certainly better attended and more interesting. A lot of people I talked to weren't very interested in today's selection of talks, and needed to get back to work. There were some presentation gems today. All in all, I think this would have been a better event if it had been a two day conference with a tighter focus.
Wednesday, 3/19
9am: I showed up to hear Bruce Momjian's talk on PostgreSQL. Bruce didn't. One of the conference organizers lobbied to get Bruce on the schedule because there are so many events that focus on MySQL, but don't have any PostgreSQL perspective. However, it seems like the lobbying was with Bruce, not with the selection committee. I still don't know why Bruce didn't show up. I hope it's a trivial matter (like missing a train, oversleeping, or just blowing the conference off).
9:45am: The next talk was from Paul Ramsey, who flew in from Victoria, BC. His company works with spatial datasets in government. I didn't follow much of his talk, because I don't deal with maps and spacial data on a regular basis. The key things to note is that the spacial data "stack" is rather intricate, and not something that lends itself to casual experimentation. Also, the number of file formats and usage scenarios is quite complex. Finally, most organizations that deal with maps and spacial data tend to be using software from ESRI. Part of that is because for many years, the only way to do any work with spacial data was using proprietary products from vendors like ESRI (actually, pretty much ESRI and no one else). Today, the open source alternatives are starting to firm up, and many government agencies working with spatial data cannot afford the thousands of dollars per seat for this software. As a result, existing shops tend to stay with their existing proprietary systems (and live with the limitations), while shops that are just staring out are much more open to using open source mapping solutions.
11am The best talk I heard all week was from Adam Rossi of PlatinumSolutions He described a typical DoD contract for a system with a lot of demands and lots of unknown requirements. The system his company developed facilitated the sharing of linguists across agencies. Lots of issues had to be resolved, including how to share data across agencies, issues of ownership and interoperability. This contract was issued post-9/11, had a limited budget and a very tight timeline.
Rossi's team used open source to develop their Java-based platform. It had to work on many different platforms, and integrate with many different base-level operating system configurations. Whatever they developed had to work everywhere. Using open source facilitated multiple redeployments (no licensing fees), integration (no requirements for a specific java app server, and their open source metaframework already handled integration with the standard app servers), and transparency (one of their customers wanted to audit every line of source they shiped). All in all, it was the story of your average open source project, but with all of the knobs turned to 11.
More mingling
2:45pm: Scott McNeil talked about the Free Standards Group, the organization that actually got the LSB to a standard status. His talk was mostly a discussion about what the FSG does and how it works to make standards for Linux.
The most interesting part of the talk was an extra slide he posted for discussion when he was finished. Scott noticed that there were a very small number of free/open source developers in 1991, and although there are more today, it's still a pretty small number. Starting in 1994 or so, a noticable number of people who directly benefit from OSS start to be noticed, and that number increased to a relatively large number of people today (no scale was provided on the Y-axis). Starting from 1994, the number of people who indirectly benefit has exploded to include pretty much anyone.
This graph was somewhat interesting, but it was clear that it was also pretty much back-of-the-napkin handwaiving. A couple of people in the audience noticed this and called Scott on it. It would have been just as easy to justify this graph using a base of commercial software developers on the bottom and leave everything else alone. Someone else in the audience (presumably an economist) noted that this relationship wasn't about economic leverage, but about commonality of interests; this same relationship can be seen with roads - there are a small number of road builders, a larger number of people who drive on a given stretch of road, and a huge number who benefit from a that road. Certainly something worth investigating further, now that the underlying economic relationship is clearer than some random handwaving...
4pm: I caught part of Jacob Hallen's talk about open source with government and industry. Mostly, his talk was about issues with software, procurement and open source in Sweden. Things are very different there, and they favor commercial software even more strongly than US policies in the public and private sectors. That is starting to change with a more agile view on iterative contracts. Unfortunately, this talk mostly sucked because of the guy in the second row who constantly asserted how these issues are well-known in the US.
4:45pm: This was probably the least fufilling presentation in the entire program. Argyn Kuketayev made a developers-level presentation of Cocoon using overhead slides. He pretty much failed each of the items on Dominus' checklist for giving presentations: reading the bullet points on the slides with very little added insight, taking great lengths to tell us what he was going to tell us, and not knowing his audience to name but three. (I'd like to see how much better a handful of perlhackers could do with presenting a Cocoon talk from scratch at YAPC with an hour's prep time.) I really don't remember much about this talk, but it does sound like Cocoon is very data-driven (all of the programming seems to be in the config file, sitemap.xml) and reasonably mature at this point. (It certainly wasn't when I last looked at it in the very early days...)
