Last week, Brendan Colloran announced Iodide, a new take on scientific collaboration and reporting that I’ve been really happy to contribute to over the past year-and-a-bit. I’ve been describing it to people I meet as kind of "glitch meets jupyter " but that doesn’t quite do it justice. I’d recommend reading Brendan’s blog post (and taking a look at our demonstration site) to get the full picture.
The great urban theorist of the twentieth century, Jane Jacobs has a quote I love:
“Old ideas can sometimes use new buildings. New ideas must use old buildings.”
Laura Thompson (an engineering director at Mozilla) has restated this wisdom in a software development context as “Build exciting things with boring technologies”.
It so happened that the server was not an area Iodide was focusing on for innovation (at least initially), so it made much, much more sense to use something proven and battle-tested for the server side deployment. I’d used Django for a number of projects at Mozilla before this one (Treeherder/Perfherder and Mission Control) and have been wildly impressed by the project’s excellent documentation, database access layer, and support for building a standardized API via the Django REST Framework add-on. Not to mention the fact that so much of Mozilla’s in-house ops and web development expertise is based around this framework (I could name off probably 5 or 6 internal business systems based around the Django stack, in addition to Treeherder), so deploying Iodide and getting help building it would be something of a known quantity.
As we move ahead, we may find places where applying new backend server technologies makes sense. Heck, maybe we’ll chose to rewrite the whole thing at some point. But to get to launch, chosing a bunch of boring, tested software for this portion of Iodide was (in my view) absolutely the right decision and I make no apologies for it.