Re: About translating documents (.xml/.sgml) in GNOME

[Christian Rose <menthos menthos com>]

tor 2003-01-30 klockan 06.51 skrev Paisa Seeluangsawat:
> I agree with Malcolm.  The biggest hassle is tracking changes in the
> master doc.  Two approaches that might help are,
> 
>   - Have a smart diff tool that can pinpoint a small change in a
>     source paragraph.
> 
>     (optional) having such tool tell what paragraph in the translation
>     the change corresponds to.  The translation file might need some
>     comment tags for this to work.
> 
>   - Using an invisible ?ML tag to break a long paragraph into smaller
>     logical blocks and do a diff block-by-block.

In practice, I don't think this becomes much of a problem (and this is
also my experience from translating Docbook XML from within po files),
mostly because very long paragraphs are rare in documentation (at least
in all the docs I've seen so far).
Documenters are usually well experienced in writing in a clear, easily
read style with not too long paragraphs, which not only makes it easy
for reading but also helps translation. Remember that we're speaking
about paragraph-sized messages, and while using po format fuzzy-matching
does not scale well, it sure works on paragraphs of the size that
usually is present in docs.
However, if some poor soul would introduce halfpage-sized paragraphs in
some doc, it sure would be nice with some sort of diff tool, but I've
yet to see the need for this in practice.


> If the ?ML in in the doc isn't too fancy, I think it should take a
> translator only 10-20 minutes to learn that markup language.  Using an
> extra tool instead of working directly on those ?ML files might
> actually make the process more complicated :oP.  Think about how
> PO-style tool would handle tables.  Working in ?ML allows a translator
> to use graphical tools to edit tables.

Oh no, please let not have this general "why use po format for docs"
discussion again... (hint: it has been reoccurring on these lists for
several years). We have arrived to a point where there is a concrete
suggestion for a conversion tool; let's not destroy that conversation by
going back to debating the basic design decisions again over and over.

If you're seriously questioning the use of po format, I suggest you try
to maintain the translation of a reasonably frequently maintained and
lengthy manual over a longer period of time. I've done that, and I'll
promise you I'll never do it again. Spending weeks of my time trying to
keep in pace with the changes in the original (chapter/title
reorginization, indentation changes, translating the exact same phrases
again and again at different places, etc). "Frustration" is an
understatement, we're talking about suicidal tendencies at that point.

If you already have that experience, I'd really like to hear about your
experiences from that. Please no "well docs are only written once", "you
only translate it once" arguments, those are not based in reality, as
far as my experience tells. Real docs written by real docs people are
customized along with the software it documents, and not left to rotten
away. Dead documents can be just as easily translated in assembly or
with a magnet on a ferromagnetic disc. It's when it comes to maintaining
and automatically accomplishing for changes in the master document the
real problems arise, and it's the same problems po format has solved for
years. We already use po format, and we use to track our status already.
Why even suggesting reinventing the wheel?

 
> Having an internationalized, easy-to-use, semi WYSIWYG ?ML editor
> might help.  Something based on Bluefish might work?
> 
> Well, I've never translated a document, so I might be wrong.

Sic.


Christian