rob.cogit8.org

DjangoCon 2009 Day 2

Here's my DjangoCon Day 2 notes...

Keynote by Ian Bicking

I would never do this justice with a short summary, but thankfully Ian posted his talk as a blog post: Toward a new self-definition for open source

Using Django in Non-Standard Ways by Eric Florenzano

This talk was a nice overview of different things you can swap out from underneath Django. At my work we use Django in pretty standard ways so I don't get to experiment with much of this but I've always known it was possible. So it's cool to hear a talk like this. Some things mentioned were: - Using Jinja2 in place of Django's templates - Not using Django's auth, with the example given being a Facebook application - Bypassing the Django ORM, with the example being getting your data from APIs - Using Django's forms in Pylons - Using the ORM in standalone scripts - Taking advantage of the plethora of WSGI apps that can wrap your Django application or scripts, with Repoze being the cited example.

WSGI middleware is something I've always meant to browse and learn more about, so hat tip to Eric for pointing out Repoze. They have many middleware components but, I agree, repoze.profile looks very useful.

Slides

The Real Time Web by Chris Wanstrath

This talk was awesome. I've played with LeafyChat a bit after DjangoDash and was very impressed. So this talk gave an insight into what the team had to figure out and create, which, after hearing this talk, is an amazing feat for 48 hours of work.

The talk was mainly about the different technologies that are currently (and not so currently) available for pushing data to the browser (vs polling for updates). The Web Sockets API is a draft proposal that is supposed to fill this space without needing to do other tricks. Refer to the slides for the 4 current methods mentioned.

Slides

It's also worth noting that server side push (or "real-time web") has been often mentioned at this conference as something that should be made more "framework-y". Coincidentally (or not?) FriendFeed has released the Tornado Web Server, which is an open source version of the scalable, non-blocking web server and tools that power FriendFeed, written in Python.

Pluggable Applications by Nowell Strite and Shawn Rider

I'm going to again defer to the slides on this one, but I like that the team at PBS is tackling a challenging problem in the reusable apps space head-on and coming up with some innovative ideas. I think we all strive for pluggable and reusable Django apps. This talk outlines some problems with the current state of reusable Django apps and how they solved it at PBS.

Slides

Making Your PostgreSQL Database Sing by Frank Wiles

What I liked in this how-to was the postgresql.conf run-down of a few of the important configuration settings, and learning about "Materialized Views", which is a table which is physically stored on disk, but is really just a view of other database tables.

Slides

I missed or didn't take notes on the rest of the great talks. See the list of talk slides on the DjangoCon wiki.

Dive into CPython Bytecode by Alex Gaynor

This was a quick (too quick) lightning talk by Alex which pointed out a module I had never heard of before: dis. What I got out of this was that looking at the bytecode of your Python code is not only interesting (and very nerdy), but can help you see what Python does with your code at a much lower level which could lead to discovering optimizations in your high level Python.

Dinner and the GitHub meetup

Dinner at Green Dragon later that night was nice. After dinner we went to Rontoms (who's site seems absolutely broken so I won't bother to link to it) for the GitHub meetup. GitHub graciously hosted this event to meet fellow GitHub users and connect with their community. For myself, this turned into a great evening of listening in on some of the great minds in our industry, and chatting with fellow GitHub and Django users.

One of the highlights was a discussion between @defunkt, @avibryant, @binarydud, and myself, @robhudson, (the bulk of which being between Chris and Avi) talking about Smalltalk and Squeek's "Image-based persistence". An idea that boggles the mind when you first hear about it, but also opens up a lot of interesting ideas. Hearing Avi talk about some of the useful things this brings to the table was fascinating.

One quick day dream I had while this talk was happening in front of me was... imagine a running Django instance that triggered some 500 error in production. Wouldn't it be quite amazing if that instance could copy and persist itself so that you could jump in and poke around, going up and down the call stack introspecting things to get an idea of what went wrong? Apparently this can't quite be done in Python -- well, maybe in PyPy. But maybe we can get the next best thing.

Only 1 day left, but it was quite epic. More to come...