Tim Blair

Flickr Workshop Review (29/09/2005)

Just over a month ago, I made the trip into London on an attempt to answer the question: Just how did Flickr do it? I attended a workshop run by Cal Henderson the lead developer of Flickr entitled "How we Built Flickr"...

A lot of people have heard of Flickr and use it — the user base has grown incredibly rapidly in the last few months alone. During the workshop Hal first described Flickr's humble beginnings as a small part of Ludicorp's "Game Neverending" — a huge online multiplayer game — to the application being snapped up and rolled under the massive Yahoo! banner.

The workshop had a very heavy technical base to it and focussed not on back-slapping Flickr for doing a great job, but more information about how the site evolved, the technical challenges of growing the site, development processes put in place by the Flickr team and problems and issues they've had to deal with on the way to greatness. It could so easily have been a "we did this, aren't we great" affair, but instead turned out to very a highly interesting and enjoyable event.

Cal's delivery was excellent, not straying from the point he was making, and rarely talking about things that wouldn't interest anyone. He covered a huge range of information in a short day, starting with the layered approach that Flickr used to develop their systems, which he dubbed the "trifle model", and moved on to cover their system environments and procedures and standards, and more importantly: why he felt they were a good thing.

Of course, Cal did come up with a couple of points which I disagreed with, such as his scale of software insanity which placed monolithic perl applications at one end and full-on OOP development with plone/zope/rails at the other, stating that either extreme was too extreme and you should aim for somewhere in the middle. I can see where he's coming from, but I think it's a case of horses-for-courses, especially where Rails is concerned — indeed, due to Rails' ease of development I felt that placing it along with huge J2EE apps was a tad unfair, and mainly the basis of my disagreement.

But that was just a single small point out of a day of information overload, the vast majority of which was interesting and thought provoking. Cal covered unicode and email problems they had and how they have been worked through; he talked about monitoring of systems and covered a lot of MySQL specific information regarding database profiling and scalability issues, including replication, which I found very interesting as it's something I've looked at on and off.

The workshop finished off with a session in a Web2.0 fashion on APIs and data sharing, talking about security and the specifics of how Flickr implemented their web services and authentication API, which allows 3rd party applications to authenticate directly through the Flickr website, so the user never types username/password combo into the 3rd party app.

I think the variety of people that attended the workshop is testament to the popularity of Flickr and just how much people look up to them in both the design and development communities. There were a number people working for themselves attending, through to the relatively small companies like ours, right up to the big players such as MTV, BBC, Multimap and the newly aquired Skype. It was interesting chatting to everyone to get their takes on the current situation of the Web2.0 buzz that's going around. It was also very interesting to get a sneak peak of the new Skype video interface!

So, my challenge now is how to come up with that next "big idea" to make my fortune...!

Article Archive (September '03 – '05)