Shipping Software

Read a great blog posting by Mark Lucovsky formerly of Microsoft and now of Google. It’s a really good read and it reminded me so something we were encountering during the 1.0.1 release of Firefox.

One of the goals of the OSL is to shorten the FOSS transaction time. That is, shorten the time it takes innovation in software to reach the end-user. We do this by focusing on what we do best; hosting, so others can focus on what they do best; innovating. We’ve done this for Gentoo, a little bit for Debian and most recently for Mozilla.

Even Firefox is still delivered in the old-school way. We build a binary release, QA, etc. and finally we push it out to distribution channel. It’s incrementally better than what Microsoft is doing today, but definitely not ideal.

On the conference call for the 1.0.1 release of Firefox, we all postulated (and I can’t believe Asa hasn’t blogged about it yet) about live, continuous updates to applications like Firefox and Thunderbird.

Mozilla is hitting a point in the FOSS life-cycle that no other application before it has. How does an open source project with finite resources distribute an application to 30 million or 50 million or even 100 million clients in a pinch? I’ll give you a hint; it’s pretty frickin’ hard.

In the beginning, we started with known software mirrors (about 10 or so) but even during the preview releases of Firefox we completely melted those down. During the 1.0 release of Firefox, we turned to the Bouncer and that has had a huge impact on our ability to serve up software but we’re still falling behind on getting the latest updates to end-users.

Wouldn’t it be fantastic to be able to continuously patch clients via XPI (yes, I know the application isn’t there yet to do it but I’m blogging — cut me some slack) and get the latest/greatest stuff on the fly? Instead of nightlies, why not have a check box under the “Advanced” menu that says “Give me the latest updates automatically” to leverage an army of testers out there. Using the (hopefully) smaller-sized XPI’s would give us a chance to get more mileage out of our mirror network.

I would liken this possibility to the fact that I have a network installer of Debian for “woody” … when I slap it into a machine and install it, right before it finishes the install it goes out to the network to see about security updates … the same would be true here. The user would download the latest Firefox and as the final process in the install they would hit UMO to get the latest patches/updates.

Firefox isn’t the last application that’s going to have this problem; Mozilla is just leading the way and finding the edges of what we can do today which helps us shape tomorrow.