mon petit blogue sans importance...

second day at pycon 2008

The second day, which really is the first day of the actual conference, was very good, with lots of very interesting talks.

The main keynote was by Guido van Rossum, about the coming changes in Py3k. He gave pretty much the same speech last year as a keynote as well, but it was still nice to hear it. Lots of good changes: i'm glad the language is evolving and that backwards compatibility is important, but not so important that the language shouldn't get better over the years. The main takeaways, however, are that:

• There is a tool, 2to3, which will be included in Py3k and will convert python source from 2.x to 3.0 for many of the backwards-incompatible changes (but not all)
• Python 2.6 will incorporate many features of Py3k, and it will be possible to write code that is mostly Py3k-safe while also running in python 2.6, except for things that generally will be caught by 2to3 anyways
• Python 2.6 will likely still be maintained 5 years from now, so there's no hurry in moving to Py3k

I'd really like the keynotes to be not just from important people in python, but from people who can really deliver an awesome presentation. Last year, Robert M. "r0ml" Lefkowitz's spectacularly delivered presentation was easily in the top 5 presentations i've seen so far in my life (there was some talk of putting it online, but sadly it seems like that was never done). I don't know what the other keynotes will be like this year, but i have yet to see someone even approaching that level in any of the talks.

Next we went to the trade floor: for the first time this year, an entire room was dedicated to vendor booths. Google, Enthought, LucasFilm (they use python for tons of stuff such as gluing all the pieces from the ILM artist's workstation to the render farm; they were super-nice, and are looking for cool python programmers to join their San Francisco office), Zenoss, Canonical (who sponsor the development of Ubuntu) and many others were there, with some schwag. The coolest schwag was an R2D2 pen from Lucas, and a pair of "sonic magnets" from WOTI. Enthought was giving out CDs of their "scientific python distribution", which for those of you who wanted to bring together all the packages necessary to do science with python know is a great idea, especially if you're stuck on Windows or OSX. Unfortunately, the OSX and Ubuntu ports aren't done. In general, all the booth personnel was technically knowledgeable and super-friendly. Rackspace is looking for people as well as customers, but even their very friendly and attractive recruiter, Erin, was able to answer my technical questions. So this wasn't E3-style booths, they were actually useful booths.

All that talking and collecting cool stuff made me late for my first talk, "Database development with Jython, SQLAlchemy, and Hibernate". I spent a few minutes before deciding it wasn't something i wanted, but i'm sure if i had been there from the beginning it would have caught my attention more.

The next talk i went to, "How Import Does Its Thing", explained ... well... how import does its thing. It was very interesting. I was unaware, for example, of sys.meta_path, which is for modules that you can import, but which are nowhere in the python path (sys.path), such as builtin modules.

Next, "Python references and practical solutions to reference-related problems" was interesting, but i did know of python's referencing paradygm before. Very educative nonetheless.

"PyTriton: building a petabyte storage system" was about some of the tech behind Mozy. While interesting, it focused solely on the tech and almost not at all on their use of python, so that was a bit out-of-place. That being said, the tech is indeed interesting, and so is their conclusion that RAID doesn't help you much at that level (or, rather, that it has too much human overhead as it forces you to replace failed disks rapidly) but their usage of Reed-Solomon coding as applied to disk array fault-tolerance across not only disks, but nodes, along with certain algorithms as to where the chunks go, has let them scale better.

I couldn't really miss "Rich UI Webapps with TurboGears 2 and Dojo", or the next 2. I use django at work and at home, but i haven't tried other frameworks very much - apart from studying them a little and trying some stuff, and some more in-depth forrays into J2EE with JBoss. This talk wasn't much about Turbogears, but it did give pretty good insight into some interesting facets of Dojo. What i liked most was (and i don't know if JQuery, which we use, can do something similar) that it has a templating parser written in javascript that lets you parse django template-formatted templates. This lets you do partial views by simply sending the data to the client, but not the fully rendered html, which saves bandwidth and server cpu. We're currently using server-rendered partial views quite heavily. We could only use this for the simplest of templates, however, as we're custom template tags, and drilling down through relationships in the django objects in our templates - 2 things we can't do client-side with something like this.

"The State of Django" by Adrian Holovaty was great, but i'm biased. He outlined the state of the queryset-refactor and newforms-admin branches, which will both be stupendously useful in many ways. The first will give us things like filtering out objects in a queryset which are dupes based on one of their fields (not based on the whole object, like the current .distinct() does). The second will give me .validate() on objects, which will allow me to finally complete a major refactor of our webservice that will make things a lot faster on object creation and mutation. o.validate() lets you check if the current field values on the object pass the different constraints you've set in the field definition in the model. At the moment, this validation can only be done by a manipulator, which expects form data, which means if you have python data, you need to stringify it into html form data so the manipulator can validate it and then convert it back to python... i offered my help in that respect a while back on the mailing list, but they figured the guy who was doing it could get it finished faster than having to help me (which is most probably correct :) ). I knew most of that anyways, but Adrian did confirm that those few changes (as well as a change in the settings.INSTALLED_APPS semantics that will make it possible to instantiate the same app multiple times for the same project) are the last big things required for them to finally call Django 1.0.

­"Django: Under the Hood­" was also very nice, and went through some of the stuff going on under the hood in Django. I already know a bit too much of that, but the talk was quite interesting.

So, overall, a great second day! It's really great to be here and see over a thousand people who work with and love python united in one place. The types of people is all over the place. I saw one dude wearing shorts and flip-flops. A few nerds wearing sandals, brown socks to the knees and saggy shorts. One guy in a suit. Most are wearing pants of course, and a t-shirt or shirt or long-sleeved shirt. It's definitely a laid-back and very open atmosphere that encourages people to mesh.

by wiswaud on 14 March 2008
Tags: chicago, django, english, geeky, python