[haiku-doc] Re: Another try at user guide translations

Humdinger a écrit :
First off, it's a bit like preparing for a race all week and then, finally at the starting line, the little fat man with the starting pistol and too tight pants wheezes "On your marks, get set...", and over track comes Vincent and shouts "Arrêtez! Arrêtez! Heureka!". The little fat man briefly considers not aiming at the sky and just to get on with it... :) :)
You got me by surprise :-p I wanted to wait for things to calm down a little on my side... But fortunately, this will be the case starting tomorrow. Anyway, nothing prevents you to start a manual translation process and then import the data in my system... It should not be too difficult.
The most important question is: Can you finish it all up and make it work from a haiku-os.org website (right, Jorge?) within the next few weeks?
If it provides PHP and MySQL (and write access to some directories from PHP), then yes, the current version should work right out of the box (with some minor tweaks if it uses another type of database engine. Install the files (12 in total), create the DB, checkout the SVN of the userguide in the reference directory, use the admin panel to add the pages, and voilà, you're done. There is one problem, however. The XML parser of some PHP versions (for instance, the one that comes with Mac OS X 10.5.7) are buggy : When you put some code into a <script> tag, it will add a <!CDATA[ around it. Browsers don't like it at all. Work around this issue is not very difficult (using some str_replace), but since this behaviour is not present in all versions, it needs to be tested first... 
Other questions:
All languages are in that one page, right?
Yes, there are only two pages : the admin page and the status pages (that show links to English and translated pages) 
If a block isn't translated, it falls back to English, right?

Yes.
Is there a summary table that shows which files are not yet fully translated?
The status pages show for instance that 10% of the parts of each document are translated in each lang. 
Is there versioning to revert changes for each language?
Unfortunately there is not, and I don't think this would be easy to do with the current DB architecture... 
How are big/small changes to the English original handled?
At this time, they are not handled. To assign translate IDs to the original document, the script just compares the contents of translatable blocks with the strings it has in its database... Of course, it is possible to match not exactly equal strings (with Levenshtein distance, for instance), but since they can contain HTML, even a small change in the string can be a big change in reality. I could add support to "fuzzy" translations though (if the original part was modified a little, don't throw away the translations but mark them as "fuzzy", until a translator revalidates it) 
Are localized images supported?
Not for now, but I see a way to do it (add another attribute to tell the JavaScript code that the "href" attribute should be translated, not the inner HTML) 
Are the finished generated languages automatically commited to SVN?
The webserver where the code is stored at this time does not allows running external programs, but adding a system('svn ci ...'); after the generation of localized files should not be too difficult. 
I'd love to have a look. Please send it over.

Should be done when you will read this email :)

Vincent

Other related posts: