Hacker News Reader

Comment by dexwiz | Hacker News Reader

Some (not Mac) users on the newest version of Firefox got lower quality Youtube video over the holidays, and the issue would would not be fixed until next year? Oh yah, next year was next week. Thank goodness, it was still fixed within the year.

Do we live in a world where this in unacceptable?

View on HN

The bug was reported on December 19 and fixed on January 7. That's at least 19 days, not just "next week". Do you think Larry Page would allow YouTube to serve 360p VP8 instead of hardware-accelerated 1080p H.264 to ~15% of Chrome users for 19+ days "because holidays"?

Well, when you consider that every bug fix carries with it the chance of introducing a worse bug, sometimes holding off on a non-critical fix until everybody's back from the holidays and able to work 100% again doesn't seem like a bad idea.

Having some % of FF 43 users (who're about 3.5% of the browser market according to netmarketshare.com) experience degraded but functional video may be better than, say, a 1% risk of breaking youtube for 100% of users and not being able to fix it for a day.

I think the point of the parent comment is that they would have done it right away even if Chrome usage had been at the same 3.5% share.

Defending Google and attacking Firefox is still the trend though, as many of the comments show here, even thus this Mozilla employe's comments are crystal clear (for once someone does that - that's great!)

> Having some [small % of users'] experience degraded but functional video may be better than, say, a 1% risk of breaking youtube for 100% of users and not being able to fix it for a day.

Not being able to roll back for a day, you mean. Worst-case you hit the undo button, which they surely have.

A rollback is a change, it can still introduce problems if thete are issues in the rollback process, a deployment freeze is a freeze even for rollbacks.

Possibly there is better communication between the YouTube and Chrome teams than between YouTube and FireFox. As in they probably cross test each release.

I run a large dev service with multiple online products, and we like most others in the industry go into a deployment freeze from about mid december to the first week in jan. Its designed to prevent new code that could cause stability issues being deployed over the holiday spell when many of our top engineers are on vacation, and our ability to handle complex problems is limited.

We (and i assume google does the same) restrict deployments to 911 fixes, which have a specific set of conditions before a fix is considered such, presumably this did not meet googles criteria. From what i read it was 6% of a minority browser traffic that was effected, which would would putt the total traffic at 1-2%, i would not break the deployment embargo for that.

Reply on news.ycombinator.com