The Debian team is always ready to join forces. The "fork" mentioned by Alain was motivated by a disagreement we had in the past about the tools to use to achieve the translation work.
Correct me if I'm wrong about this: one of the effects of the "fork" is that Debian has in this instance ("man-pages") created a translation effort whose results are used only by Debian. That seems unfortunate.
Yes....and no.
From what Nicolas François told me, he always pushes updates done by the French team back to Alain.
What Alain does with them is unknown to me. Maybe these updates are part of the extra work leading him to burn out which is understandable for me.
That explains why I think we should try to find good solutions that would make all contributors happy, despite the disagreement we have (or had). We are all involved in FLOS for long time enough to know that dispersed efforts are the best way to discourage contributors.
This is why the po4a tool was developed in the past, by Martin Quinson (member, at that time, of the Debian l10n French team). That tool has been used since then for more and more manpages i18n for Debians-pecific tools (dpkg, aptitude, etc.)
(Of course, I have no experience of translating man pages :-). I can guess at some virtues of po4a, but it's just a guess.)
Amon the virtues I see and often "mets en évidence":
- work a format (PO) which is familiar to translators and is easy to learn and deal with (there are good GUI tools to work with gettext files)
- guarantee that the translated manpage is always up-to-date. This is IMHO the killer feature. po4a builds translated manpages that are exact counterparts of the original files. When the translation is incomplete (for instance, an untranslated, or fuzzy, paragraph), the English part is put in place. One can set a threshold (usually 80%) where the translated document is no longer built, to avoid too incomplete translations. This allows a document to be still translated and still have the entire original information, even though a minor change in the original version has not been coped with by the translator
*This* kills the most often used argument *against* translated manpages: "they are never as up-to-date as original versions are" which usually turns into the geek motto: "je n'utilise jamais les traductions car c'est jamais à jour".....j'entends les geeks francophones utiliser cela à longueur de journées.
The argument *against* using gettext-based translations is often that the context might be harder to guess (gettext-based translation is done on a string-by-string basis where one string is one paragraph).
This is indeed true and probably the minor price to pay for much easier access to new translators (working directly in *roff files is a real PITA for non geeks (and many people interested in helping with l10n work are non geeks).
So, as of now, the French translation of manpages done in Debian is done with PO files, then the files are regenerated *and the changes are sent to Alain* by Nicolas François who is currently doing the huge work of recollecting all this.
(I'm not so sure that the changes have always actively and regularly been sent back to Alain, but of course Nekral and Alain know more about this.)
Well, Nekral is just like many people with great skills and great motivations: he's doing a lot of things (in Debian and outside Debian), so he might very well have missed some exchanges.
I understand that, doing so, we're pushing towards the adoption of gettext as preferred format for Linux manpages i18n/l10n....and this might hurt some people who are doing that work manually or semi-manually for years. But, really, after an itial adaptation time, that's worth considering it. If some of you have the chance of running Debian or Ubuntu, just try looking at the dpkg manpages...:-)
For Alain, that may sound like beating an old horse as we had such discussion a few years ago, already. However, maybe are we more ready now than we were at that time. It's worth considering this, IMHO.
I know that Alain has been doing a lot of work recently to improve his translation by reviewing all of the diffs with Debian's translation. I guess that this has the effect of decreasing the differences between the two translations. Now hypothetically speaking, suppose the upstream maintainer switched to using po4a. (I know that Alain resists this, but let's just explore possibilities here.) In that case, if everyone wants to move to a single transation, I suspect that at least some team's work needs to be discarded. If I understand things right the only choices are:
a) Simply switch over to using the current Debian translation, and make that "upstream". This would have the effect of discarding Alain's work as embodied in the differences between the two translations. Disadvantage:
- I think this would be a tough path for Alain to agree to, since he
has invested so much effort (perhaps evern more than the whole Debian team combined, but that's just *my* guess).
This is certainly not the way to go, for the exact reasons you're giving.
b) Convert the current upstream French pages to po4a, and the Debian teams switches to working on that version of the translation. This would have the effect of discarding the Debian team's work as embodied in the differences between the two translations. Advantages:
- Acknowledges Alain's tremendous work, and is a clear demonstration
of goodwill by the Debian team. Disadvantages: *Alain must switch to po4a; but perhaps in the end it is worth doing that.
- Some of Debian's work is lost -- but maybe the loss is not enormous,
because Alain has been doing a lot of work comparing the two translations and changing his. (Of course, you can deduce better than me the magnitude of the current differences between the two translations.)
Actually, the Debian work would not be lost.
At the beginning, we would have two sets of PO files: - those converted from Alain's work - those in Debian
Then, file by file, both "parties" would need to examine the diff and do what's needed to reduce it to zero, if possible.
If there's a disagreement, then the Debian package can still keep the diff (quilt makes this fairly easy to do). There could even be cases where a Debian-specific diff could be well worth it even if there is *no* disagreement (think about the behaviour of a given utility being different in Debian....we have such cases in shadow, for instance)
c) Systematically review each and every one of the 850 pages in man-pages, to remove all of the diffs, taking whichever is the better of the two translations in each case, in order to arrive at a single translation that everyone agrees on and can go forward with. Advantage:
- this is "the perfect solution" -- everyone arrives at a translation
that they agree on. Disadvantages:
- The perfect is the enemy of the good: this approach is a *huge*
amount of work, I guess, and would take quite some time to complete. In the meantime, I'm marching on, producing several new man pages each week, and changing many others.
Yes, which is why I think it wouldn't work
- Alain must switch to po4a; but perhaps in the end it is worth doing that.
Are there any possibilities other than those I've listed above?
And I can't resist asking: which of these options could each side consider as viable?
I think that b) is the most reasonable, indeed. I fully understand that, saying so, we're saying "switch to PO" to Alain....:-)