Closed Bug 1637549 Opened 4 years ago Closed 4 years ago

update download page for macos < 10.12 moving to esr

Categories

(Webtools :: Bouncer, task)

task

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: jcristau, Assigned: oremj)

References

Details

We're looking at removing macos < 10.12 support from release Firefox, and moving that population over to the esr channel with 78. I guess this will need some updates to our download page.

Blocks: 1637533

I'm guessing we'll need to update the download page/buttons to care about the MacOS version. What do you think Alex? Do we already do this for some things?

Flags: needinfo?(agibson)

When we did this for Windows XP users (serving them ESR), I believe we put this type of logic into Bouncer, as opposed to bedrock? If I had a preference for consistency, I'd say we should probably follow suit here. Thoughts?

Flags: needinfo?(agibson) → needinfo?(pmac)

That would do the most to ensure that people didn't download the wrong version for their system. If the bouncer team will take it that would make sense to me. If not then I'm sure we could do something if need be.

Flags: needinfo?(pmac)

Justin is this something we can do in bouncer? If so I guess we can close this one.

Flags: needinfo?(bugspam.Callek)

I think in the XP/Vista case we had some magic support already special cased due to the need for SHA1 installer details, that we don't have in the OSX 10.9/10.10/10.11 case we're talking about here, so I don't know if/how we'd do this solely on the bouncer side.

Prior discussions in Bug 1339226 and Bug 1340263.

I'll reach out with :rail and :oremj who did this work last time to validate my presumptions here. [Releng no longer has direct access to bouncer UX]

Flags: needinfo?(bugspam.Callek)
Flags: needinfo?(rail)
Flags: needinfo?(oremj)

Yeah, bouncer handles the WInXP special case in here: https://github.com/mozilla-services/go-bouncer/blob/eff581420e089c99b5e9f950091ee53f8624dc0c/handlers.go#L365-L371

I guess we can do something similar for macos < 10.12 if we can sniff the OS version from the UserAgent field.

Flags: needinfo?(rail)
Flags: needinfo?(oremj)
Blocks: esr78

Actually looking into this more, it looks like Bouncer is not likely where we want to do this.

Bouncer was for the sha1 requirement for XP so we needed actually different binaries for those users.

However in Bedrock we do the actual work of pointing the users at ESR for XP/Vista:
https://github.com/mozilla/bedrock/blob/7d96eecdd4a11cbb057ec1153f40b76eeec5b875/media/js/base/site.js#L157

More specifically that was this change https://github.com/mozilla/bedrock/pull/6044

However in Bedrock we do the actual work of pointing the users at ESR for XP/Vista:

I don't believe this is strictly correct. PR https://github.com/mozilla/bedrock/pull/6044 added a warning message to users on XP/Vista, however it didn't do anything to redirect to the most appropriate download. I'm not sure where this logic lives currently, but I don't believe it's in bedrock.

Bug 1340263 seems to be a good paper trail here for what was originally done. If we really need to, we can certainly try and handle this in bedrock. But if we are still already doing this type of redirection in bouncer, then I think keeping it all in one place probably makes more sense.

After discussion in slack yesterday, Oremj was going to help with getting this setup in bouncer.

Component: Pages & Content → Bouncer
Product: www.mozilla.org → Webtools
Version: Production → other
Depends on: 1649205
Assignee: nobody → oremj
Status: NEW → RESOLVED
Closed: 4 years ago
Resolution: --- → FIXED

Verified using macOS 10.9/10.10/10.11 as follows:

Status: RESOLVED → VERIFIED

It makes sense to drop support for macOS 10.9 and 10.10, as there is very few reasons to stay on 10.9 (old macOS design) and 10.10 (SIP not present). but there is no point to drop support for macOS 10.11 considering it is the last version of macOS without SSE4.1 requierement, there are much computers stuck in this version officially and unofficially.

See Also: → 1840562
You need to log in before you can comment on or make changes to this bug.