Credit where credit is due: half of this tutorial is by James Bennett (http://b-list.org/).

Great work.

Is there any audio/video available from the PyCon presentation?

FWIW, not paying attention to the date I didn't realize PyCon was in-progress. If memory serves, videos usually don't end up online till after PyCon has completed, though I could be wrong.

Exactly - the conference is being taped, but videos won't go up online until after things wind down.

I came across some of the uploaded videos on blip.tv

URL?

It seems django would benefit by brining some of this "in-house", either officially or through blessed third-party tools in the same way that rails does.

If deployment and testing is something that everyone has to do in a very similar way should it not be a problem that is solved by the framework the same way that sessions are?

The thing is that actually very little of this is unique to Django. The types of tools you need to deploy dynamic-language-based web applications don't differ depending what the "P" in LAMP stands for.

That maybe true (Capistrano which you reference was pulled out of 37 signals) but I think my general criticism still stands (+ standardization in and of itself has benefits in that it helps people help each other).

The rails community from the beginning has had a more holitstic view of the problem it is solving.

The third tab on the rails site is about deployment. There are three performance monitoring companies and 6 different rails specific hosts.

Other examples: db schema evolution (migrations) and rake tasks. There is also a litany of useful maintance tools you can find on github.

This is not just in reference to code either, its everything else too, including documentation.

I like django, I prefer django, but in this one critical area, the last mile, rails does better. And its frustrating because its not a hard problem.

sessions are in contrib, not part of the framework.

Some of the testing stuff is "in-house".

It's not even close to true that everyone's deployment is (or should be) very similar. It's a bit more arguable that everyone's testing should be similar. But forcing one way is not the Django way and one of the big reasons it is loved.

amazing. I'd love to see the lecture that this accompanied.

This document is covered with "FAIL" stamps, but they didn't even mention HAProxy as a load balancer! Anyway, I think that slide 135 is what most people do, maybe adding a DB cluster.