On Fri, May 12, 2006 at 12:23:57PM +0200, Michael Van Canneyt wrote: > On Fri, 12 May 2006, Marco Ciampa wrote: > >On Fri, May 12, 2006 at 11:23:23AM +0200, Michael Van Canneyt wrote: > >>On Fri, 12 May 2006, Marco Ciampa wrote: > >>>On Thu, May 11, 2006 at 03:33:02PM +0200, Michael Van Canneyt wrote: > >>>>I think the po2xml is just used to translate the program strings which > >>>>appear in the docs (used to refer to button captions etc), not for the > >>>>actual > >>>>documentation text. > >>>Wrong. It is used for documentation writing too as you can see here, > >>>for the LinuxFromScratch manual: > >>> > >>>http://www.mail-archive.com/lfs-dev@linuxfromscratch.org/msg06863.html > >> > >>In that case, I understand even less why they use it. In my opinion, > >>the .po format is totally unsuitable for this kind of thing. > >Sorry, I do not understand and I really want to see your point. > >Could you please tell me why you consider .po files not suitable for the > >job? > >I do not know of some other tool able to manage the update problem of a > >multilingual translation as the gettext tool. May be I'm wrong but I > >think that such a tool simply does not exist! > > po files are designed/meant for captions, short one-line descriptions to be > displayed in a program, produced by gettext from sources. Right. It is the kind of use I intended for gettext. We both agree about po files scope but I think you missed the point.
> Gettext uses a very inefficient algorithm for translation: it searches > in the .po (or .mo) file for the original text, and then returns the > translated text. You can't get more inefficient than that. No. It is really quick and it doesn't search the .po file but its compiled form, the .mo and anyway, I never intended to use it as a final product, so I do not really care if it's fast or not. > Conclusion: > The .po format is not designed/meant for large pieces of continuous text. Yes, I agree again and the use I have in mind ([EMAIL PROTECTED] in this ml grasped the concept quickly, see: ---- > Maybe a "source format" file combination of "*.po" files and a XML > file: > > -------- > <xml> > <title path= "title.po" /> > > <contents path= "contents.po" /> > > </xml> > -------- yes, sort of... ---- ) is to use the interesting propriety of the .po files with the gettext utility suite only for source code, not for the final result. Infact I never inteded it to. A paragraph is always a short text, in general a fiew lines, so it's ideal for gettext & .po strings. > I agree with you that a good tool for translations does not exist. > The question is whether such a tool should exist. The right tool do not exist but gettext is the most andvanced tool in this direction thought. > A simple diff > (with nice gui) in any text editor can help you in keeping your > translation up to date just as much... No, really not, it's a nightmare, believe me! My main occupation (as a volunteer) is to translate free software programs and manuals (like GIMP) and I can assure you: without gettext revision control it's really difficult to keep track of the modifications inserted at random (in time and location) by contributors even using a language (the english) as a reference and with cvs (and cvs web interface). > There is no need to use a .po file format for this. For the final product, no. I really intended to obtain html _AND_ any other suitable for browse/search format like chm or pdf or info or whatever you like. > How would you translate a document written in Latex, or Ms-Word or > OpenOffice Writer? Cut it in paragraps and paste it in many .po strings and the work is easier than that with any other source format. > There are no good tools for this, and I don't > believe it's possible to write such a tool other than some visual > aid for detecting differences... Have you used tools like intltool-update from the gnome project lately? It works like a sharm! Is it possible to obtain the same result with any other tool? Do you know of any similar tools? Without gettext utilities like msgmerge? > For example: > The fpdoc format is much more suitable for translation, because all > pieces of text are identified with unique names. The text is really a unique identifier, or not? > It would be much more easy to write a translator program for that. Yes its simple in a one-shot effort. Pray to be not the manual mantainer...and why should you reinvent the wheel when there are so many instruments like poedit, kbabel, emacs po-mode, even rosetta web interface from the ubuntu project!?! Remember: the problem is _not_ the easyness to translate the manual _but_ the maintainability of it! -- Marco Ciampa +--------------------+ | Linux User #78271 | | FSFE fellow #364 | +--------------------+ _________________________________________________________________ To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe" as the Subject archives at http://www.lazarus.freepascal.org/mailarchives