Updating translation workflow

[Warren Block]

On Tue, 20 Aug 2013, Gabor Kovesdan wrote:

> Em 20-08-2013 19:10, Warren Block escreveu:
>> On Tue, 20 Aug 2013, Gabor Pali wrote:
>> 
>>> On Tue, Aug 20, 2013 at 4:13 PM, Warren Block <wblock at wonkity.com> wrote:
>>>> Well, yes, but this appears to be manual and does not automatically
>>>> translate the same strings that are found in other documents. That is 
>>>> part
>>>> of what the newer automated systems do.
>>> 
>>> Erm, I am a bit skeptic about this as nouns (and expressions) in
>>> Hungarian may be conjugated; inserting them into a sentence would
>>> require one to understand the original context of the word -- that is,
>>> if some automated system could translate those words in case of such a
>>> language, no humans would be required any more at all... :-)
>> 
>> I would think translations of larger strings ("This is an example of a 
>> FreeBSD system") would override translations of smaller strings 
>> ("FreeBSD").
>> 
>> But it's not clear, and a test would help.  Certainly it can be made to 
>> work.  Other large projects are using these systems, with a lot of content 
>> translated.  Because of the translation database, rewrites or massive edits 
>> would force the translators to start over.

> I've worked with Trados before (commercial CAT software) and there you can 
> set up a match percentage limit. If the limit is met, you are offered the 
> translation of the matching text but you are able to edit it. From the 
> translations you write a dictionary file is created, which stores earlier 
> translations, so as you progress, there is always a wider basis of 
> translations to search for matches. I think using similar methods in FreeBSD 
> would be very practical if we can find proper open source CAT tools.

As I understand it (imagine it, really), this is how Pootle works.

> I agree with Warren that we should improve our workflow since it is 
> really outdated. As for expanding the entities, I'm concerned that we 
> may loose functionality so I'd prefer just handling entity references 
> as plain text there and translating them in the entity files or 
> something similar.

I think you mean expanding not-so-simple entities like &man.ls.1; would 
lose the functionality of it making a link to the man page.  I agree, I 
was only thinking of simple entities like &os;.

If entities were passed through unchanged, the translator would have the 
freedom to leave them unchanged or convert them to plain text if needed. 
But this means we need some other change to itstool, which currently 
chokes on entities.  The older xml2po has an option to keep entities:
   xml2po -k article.xml