Firefox 34 Post mortem
- (lmandel) Pushing out the release a week allowed us to take additional stability fixes. We had the flexibility to do this in 34 because we had previously added a week to the 35 cycle. I don't know how easy it will be to do this again. Those extra fixes were also good because we had less developer time for stability/graphics issues since many people were still working on 33 point releases.
- (lmandel) releng and QE did a great job of supporting multiple releases (34.0, 34.0.5) the same day
- (lmandel) releng figured out how to throttle differently for different locales. This may be useful in the future.
- (lmandel) Releasing on Dec 1 was pretty smooth even with travel and limited availability.
- (hwine) earlier explicit timeline would be helpful - 0500 vs 0800 PT release was up in air until Sunday
- (liz) hal's timeline combined with my etherpad compilation of what we were going to do and when (extracted from emails) was helpful. It was difficult to figure out how to test and handle 2 releases at once.
- (lmandel) Late UX change required on mobile. I'm going to look into a UX review when we merge to Aurora to try and prevent this from happening again.
- (lmandel) There was some user confusion about 34.0 and 34.0.5 and which release they should be on.
- (lmandel) Lots of feedback about landing a half baked search UI change so late in beta.
- (lmandel) Lots of user confusion (internal and external) about Hello icon not appearing because this was a throttled feature release.
- (bsmedberg) - The Hello icon wasn't even present in the customize panel, which was the most confusing bit. (We announced a feature without making it available to everyone and without explaining why it wasn't available to everyone yet in the UI)
- (lmandel) marketing suggested not promoting features that are being throttled.
(cww) Confounding issues: Flashblock update broke videos. (cww) Flash blocking basically at the same time as unthrottling. - (bsmedberg) This was partly because we unthrottled the same day as MS patch Tuesday which revealed an active exploit.
(cww) We had a hard time interpreting data on search feedback because the tight throttling meant numbers were less than expected but feedback was high.
- (lmandel) Late Android L changes meant added risk to the mobile release. Can we do better to identify issues earlier with the next Android release?
- (tyler/cww) Primary user issues with release are flash (above) and search UI changes.
- had to scramble, things we could have done better if we had more time
- always good to know about different release mechanics as soon as possible
- would have found more issues earlier if included more QE