L10n opt-in wondering

[Cédric Corazza]

Hi,
I've been wondering for several days about the usefulness of l10n
opt-ins for l10n betas and RCs releases on this newsgroup:
- this causes 30+ posts on the newsgroup in the yet overwhelming
information world of Mozilla
- check-ins impacting l10n land after locales opt-in , trusting their
green tinderboxes
In my humble opinion, l10n teams just need to be informed of the
deadline for their locale to be green. If it is the case, they're in, if
not, they're out.

Thoughts ?

[Alexandru Szasz]

I asked the same question and the answer was that some teams may not
want to release a beta into the wild.

But to reduce the unneeded message flow, I or someone else could
create a small php application on l10n server for the opt in process.

--
Alexandru Szasz

[Ricardo Palomares Martí­nez]

Cédric Corazza escribió:

> Hi,
> I've been wondering for several days about the usefulness of l10n

> opt-ins for l10n betas and RCs releases on this newsgroup: (...)

> In my humble opinion, l10n teams just need to be informed of the
> deadline for their locale to be green. If it is the case, they're in, if
> not, they're out.
>
> Thoughts ?

I agree with you, but I wonder if such strategy would mean more work
for Axel and other Mozilla staff. :-?

Regards.

[Axel Hecht]

I gave this post a few days to collect input from other localizers, told
Cedric so up front. Here's my take:

Our automated testing infrastructure is really poor. Like, all we do is
build it, and run some source checks on it. Which are far from
extensive. They're better than what we did for 1.5, but far from extensive.

We don't have any automated runtime testing, for example. We're in the
process of getting that these days, but as of now, we don't have it.

So, in order to actually ship something that's OK, we need to make sure
that someone actually took the build, launched it, and did some level of
smoketesting with it. On hopefully all three platforms.

Taking Cedric's proposal, we'd have to rely on localizers to actually
intentionally regress their builds to fail one of the source level
checks if they found a blocker regression in their build, without the
time to fix it. That sounds bad, too.

We're still struggling with that approval thing, I don't like it either.
But to me, it seems to be a fair compromise for now.

In my dreams, we'd have a web-app with ldap authentication where you
could log-in, and just flip a switch and be opt-in. But those are
dreams. Gonna file a bug on that anyway, as part of my dashboard buglist.

HTH

Axel