Thunderbird Discussions with Mozilla (was Marketing Module)

Andrew Sutherland asutherland at asutherland.org
Fri Jan 9 03:01:10 UTC 2015


On Thu, Jan 8, 2015, at 07:28 PM, Joshua Cranmer 🐧 wrote:
> I mentioned that I was sad that some of these things needed to be 
> explicit, and one of the main reasons is that you would think that 
> common courtesy would suggest that they should be done.

Some of the A-F points are requesting things that go far beyond common
courtesy, however.  Points C, D, E, and F are all imposing required
actions on developers working in mozilla-central, although only point E
includes the threat of backouts.  (And the awareness goals of D and E
seem like they can already be met by the Thunderbird developers watching
the affected bugzilla components.)

A thought exercise for common courtesy in this case is to imagine if
Thunderbird was a new project that didn't previously exist and started
developing against mozilla-central (let's say Newbird) and determine
what would be reasonable in that case.  Just like developers who show up
and are interested in using XULRunner native C++ stuff using internal
APIs, Newbird would be strongly advised against it and told it was
unsupported.  No sane person would suggest Newbird use mozilla-central's
build system internals! ;)  And I think one could assert it would be
unreasonable for Newbird to expect that its trunk would always be able
to build against trunk mozilla-central.  This is both a silly thought
exercise and an unfortunately realistic one.

The reality is that Thunderbird had trouble keeping its builds green
even when mozilla-central development was proceeding much more slowly
and MoCo/MoMo were funding full-time engineers.  A lot of the problems
you enumerate seem to have to do with the goal of keeping trunks
building together.  While I understand why Thunderbird would want this,
it's not clear that it's reasonable or a good idea.  It seems like
Thunderbird could just use specific revisions of mozilla-central that
get bumped periodically; much like a vendor branch.  I think this has
been previously proposed?

I want to make it clear that I know how hard it can be to be a
Thunderbird developer; I was one for a while!  But it's also very hard
to be a Firefox/Gecko developer.  The examples of build system changes
and load group changes are excellent examples of significant
undertakings that require coordination within a large and not always
well-documented code-base and that are likely personally exhausting. 
And they yield massive benefits; the "mach" build tool and Firefox build
system improvements have made Firefox development significantly less
painful.  Requiring that Thunderbird be 100% considered under pain of
backout potentially makes a hard job impossible or may just be
sufficiently demotivating that engineers avoid improving
mozilla-central, which is something I think none of us want.

Andrew



More information about the tb-planning mailing list