On: How is WordPress.com made?

I help make WordPress.com. Any questions?

Can you talk a bit about how you scale it? I'm especially interested in if you use a shared filesystem like NFS to keep the blogs.dir consistent across multiple web front ends, or do you use something like fuse on top of mogilefs or HDFS? Do you use varnish? Could you share your config?

Thanks a lot!

We use a different domain for uploaded files. Example: http://andy.files.wordpress.com/2011/01/walkingdesk-e1294433... This way we can direct those requests to services better designed for serving and caching static files. We use NFS for blogs.dir (iirc) but sparingly.

A good place to start would be Barry's blog: http://barry.wordpress.com/

Hes not updated it in quite awhile though.

Barry has the strength of ten sysadmins.

I'm curious if wordpress devs ever contemplate doing a complete re-write of the codebase. It has one of the worst "architected" codebases of any major open source product out there and even people in #wordpress-dev on freenode make jokes about it occasionally.

Contemplate, sure. Seriously think about, no. It's too hard to make the case that it's an economically viable activity.

I was working on a blog addon service that required adding a small javascript snippet to the page to add a button on the side (like GetSatisfaction or UserVoice) and found that this was possible on almost any service except WordPress.com.

I'm well aware that letting users add js to their sites is a major security issue, but it seems like something that would also have a lot of upsides.

How did buddypress come about?

http://buddypress.org/about/story/

Andy Peatling invented it before joining Automattic.