Fwd: Tx 0.7/0.8 question

[Paul Frields]

Dimitris gave me permission to forward this to the list -- a followup
on my action item from last week's meeting regarding the Transifex
migration.

Paul


---------- Forwarded message ----------
From: Dimitris Glezos <dimitris at glezos.com>
Date: Thu, Feb 4, 2010 at 3:56 AM
Subject: Re: Tx 0.7/0.8 question
To: "Paul W. Frields" <stickster at gmail.com>


On Thu, Feb 4, 2010 at 3:17 AM, Paul W. Frields <stickster at gmail.com> wrote:
> In a little over a month, March 11, the Docs group will be approaching
> a deadline for their guides.  By that time, they need translate.fp.o
> running Tx 0.7.  There's a ticket open for that, and work is
> progressing:
>
> https://fedorahosted.org/fedora-infrastructure/ticket/1455
>
> I asked in tonight's Docs meeting about that deadline, and whether the
> Docs team had considered any alternatives.  One alternative is to
> continue using the old system, with bad (and tedious) hacking required
> on the part of Docs maintainers to split/merge Publican POT files for
> use on the older Tx.

This is a bit hairy and we should probably avoid it, like you implied.
We need to channel resources into more creative and important stuff.

> Another alternative might be using transifex.net for one release.  As
> transifex.net is a completely free software platform, doing this would
> not go against the Fedora mission or methodologies.  However, it would
> depend heavily on the portability of 0.8 data/schema back to 0.7.  How
> hard is that, and is this viable at all if it comes to an emergency?

0.8-alpha is out, so Fedora could upgrade straight to 0.8 in a few
weeks anyway. Additionally, the work needed to re-create a few
projects on translate.fpo is very small -- just a few clicks away. For
these reasons I wouldn't worry too much about migration. =)

We'd be happy to help in any way we can. Adding Docs on Txn should be
a few minutes' work. There's a ticket open to enable Transifex.net
submit to Fedora Hosted:

 https://fedorahosted.org/fedora-infrastructure/ticket/1687

For the record, Transifex offers migrations both in a forward way as
well as backwards. We haven't done heavy tests on the latter, but they
should work.

> Just to be clear, I have full faith in our Infrastructure team;
> they're true genii.  But it's always good to have a backup plan.

I too believe we have an excellent Infrastructure team. Our challenge
is resources, since we don't have people actively maintaining and
pushing our L10n infrastructure. We could try this out and see how it
goes. It's how we first tested Transifex itself, right? First it was
docs, then we moved more projects.

By the way -- we could consider switching to transifex.net in general.
We roll out incremental upgrades there and, as a plus, using an
upstream instance will help in many areas. If you consider projects
like Pulseaudio and Packagekit, this makes total sense. If you
remember, we have had complaints in the past about
system-config-printer. Even for projects like Anaconda it makes sense:
Moblin is using Anaconda and it's a pity to lose translators. Examples
of projects using a hosted free software are Qt and Maemo using
Gitorious (qt.gitorious.org and maemo.gitorious.org). Even Moblin is
considering to switch to something like moblin.translate.org. This way
they can channel resources in competitive advantages instead of
Infrastructure.

In any way, this should be a separate discussion. If we would have
"good multilingual support" a primary goal of us, we should start
thinking more strategically about how we are pursuing it, why
Launchpad is investing so heavily in Rosetta (Launchpad's L10n tool)
etc. Just some food for thought. =)

-d



--
Dimitris Glezos

Transifex: The Multilingual Publishing Revolution
http://www.transifex.net/ -- http://www.indifex.com/