Rosetta/bzr (was: What is an Upstream?)

[Tim Morley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 7 Nov 2006, at 23:39, kiko at async.com.br wrote:

> The truth is, though, that we have really good infrastructure but only
> average-power tools to manipulate the data that is there. This will
> change over the coming months as we focus on tools to make QA and
> collaboration much easier.

Well that's certainly great to hear. My knowledge of Rosetta is  
mostly limited to the user's perspective, so it's good to know  
there's more to the back end than might be inferred from the visible  
front end.

>> Tim:
>>    people who wouldn't otherwise have been able or been bothered;  
>> but this
>>    came at the price of now having no idea who translated what,  
>> having no
>>    accompanying notes saying e.g. "This translation is just a best  
>> guess;
>>    feel free to suggest something different", or alternatively,  
>> "Here is a
>>    list of quality references backing up this translation", so we  
>> now have to
>>    go through and double-check everything, even things that the  
>> original
>>    translator has already researched and checked in detail,  
>> because we've no
>>    way of *knowing* that any given entry has already been  
>> researched and
>>    checked.
>
> Christian:
> So from this paragraph, for a checklist of features that you'd like to
> be able to better manage translations:
>
>     - Who contributed an approved translation (implemented)

Yes, definitely; I believe this information is already recorded, but  
I need to be able to use it too.

>     - Who approved a contributed translations (to be rolled out  
> next week)

Yes.

>     - Whether the translation should be reviewed or not (implemented)

Well, I'm not too bothered about a binary setting that says "Needs  
review/doesn't need review", although it would have its uses. What  
I'm more concerned about is the next item:

>     - Translator notes/whiteboard for each string (planned but not
>       implemented)

Yes. Absolutely. Definitely.  :o)  Now, before we used Rosetta, we  
managed quite adequately by just adding comment lines to the .po file  
using whatever .po editor we liked (and to be honest I was a bit  
shocked that this wasn't implemented in Rosetta). We rigorously added  
author's name and date after each comment, and some strings had half  
a dozen or more comments above them, through which we could trace the  
thought process that had brought us to the current translation.

Thus, on finding a translation that looks potentially suspect, any of  
us could immediately see whether it needed attention or not. If it  
had no comments at all, and it looked dubious, it probably needed  
attention. If it had a comment saying, "Dunno what this means  
really", then it *definitely* needed attention. But if it had a whole  
conversation thread between two or three experienced members of the  
team, who'd reach an agreement, then we could see that in a matter of  
seconds, and move on to something else.

That's important! What we're having to do with the .po files we've  
got from Rosetta is to go through and retro-fit all such comments as  
we proof-read, and it's taking bloody ages to do from memory;  
basically we're having to repeat all sorts of research to re-discover  
why we chose a particular word or phrase.

> For the "already checked" part I've got a question. Currently the way
> Rosetta works is, if you are using moderated mode, that anybody can
> contribute a translation, but it won't actually be used until somebody
> who is in the translation team approves it. We know there are
> limitations and confusions that stem from the fact that the  
> translation
> team isn't clearly a review team, but if that was solved, do you think
> we'd still have this sort of problem? In other words, is the fact  
> that a
> translation has been approved by a reviewer (and you know who he is)
> enough for you to be happy that it has been properly verified?

Without being able to see the thought process leading to a particular  
translation, no it's not enough. It's useful information, to know  
that Mr X or Miss Y thinks a translation is OK, but it's not sufficient.

Polishing the translation is very much a process, not a one-step  
operation -- one big grey area, if you prefer -- and the current  
Rosetta implementation seems to work on the assumption that any given  
translation string is either correct or incorrect, and once somebody  
has declared it to be correct, we never need to look at it again.

> Perhaps that in addition to a whiteboard?

Definitely, even if the "whiteboard" just takes the form of comments  
in the .po file. That in fact *would* be sufficient (and would have  
the advantage that exported .po files could be made to contain all  
the translators' notes, as comments, if required).

Thanks for taking the time to plough through my message, extract the  
critical points, and come back and ask for more.  :o)

Cheers.


Tim

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFFUSFi4qQfaYPSZYwRAhGSAKCPPHGnfOxEpUr8RTXes4qw9sMrHwCfZ5K2
6hS5fxtlaQIcKBobbcpKuR0=
=Fpmi
-----END PGP SIGNATURE-----