Le dimanche 30 octobre 2005 à 19:11 +0100, Giuseppe Bilotta a écrit :
> Hi Nicolas,
>
> I'm glad you took your time to describe the scenario I
> couldn't understand. Now I see how a situation like the one
> you present can create horrible problems with the current
> implementation.
Well, you did it first by explaining your ideas which enabled me to
understand where we diverged.
> And I think the solution I present would solve that kind of
> problems *too*.
Not so sure :(
> the 'stupid
> & dangerous'argument would mean you can't give them pens or
> scissors because they can poke their eyes out.
The difference being software is supposed to be smarter than scissors,
and when an application lets users do obviously stupid things they blame
the stupid application (except when it has been elevated to fact of life
like MSO)
> But this is a
> totally different issue, let us see how my proposal would
> help in your case.
Agree
> So you have these beautiful multilingual documents. They
> have been translated, and the translators have tagged the
> various parts with the appropriate languages. Please note
> that whatever the active method is (yours or mine) we *have*
> to assume that translators *will* tag the text correctly,
> ok?
Yes, if they don't no one will. But they need spellchecking to help them
so they won't forget this step.
> So we can as we assume they do it by my method: they
> create a bunch of styles ("French_language",
> "English_language") that only set that property, and tag
> the text accordingly. With your method they would have this
> other method of tagging the text and tag it, same thing (it
> is not in this process that my method and yours are
> different)
>
> Problems begin when you pass the document back to the
> typesetting team, and in their rearrangements, restyling and
> what not they inadvertently change the language in the
> Default style to Turkish.
>
> What happens to your beloved documents? NOTHING! Why?
> because the actual language used is the one of the topmost
> style layer ("French_language").
>
> I assume your typesetters know they shouldn't alter the text
> (delete parts, add gibberish, white it out so that all pages
> come out blank, etc), correct? Well, just the same way you
> can tell them to not touch the Whatever_language styles.
>
> If necessary, it could be possible to add extra checks like:
> ensuring the language styles are always toplevel, locking
> them or whatever, but in the end you cannot create any
> mechanism that is totally foolproof.
Agree there too except so far no safeguards have been proposed at all.
But they could. I almost proposed several but then I remembered my
typesetting days (I'm sorry, I'm a dangerous individual who got exposed
to translators and typesetters).
You're assuming there there will always be a 1<->1 relationship between
the tagging translators use and the tagging typesetters use. This is not
the case. As you will know if you've done any professional typesetting a
colour page costs a lot more than a black and white one. So I can
imagine easily typesetters deciding to turn a few special pages to
colour, and leave the rest black and white (there are probably many
reasons why one may format parts of a document differently than others,
this is only one scenario).
Special formatting may include changing text colour based on language,
using colour language icons instead of B&W ones, etc.
Now if this was something planned from the beginning, your scenario
would still work because the translators would have been handled
sections with different set of styles and been told to use greek and
colour_greek depending on the place in the document.
Unfortunately this kind of decision is left to the last moment (going
full colour, B&W only or mixed depends on your budget). If you had
something like my conditional styling all would be ok. But in your case
since you depend on language<->format relations staying the same during
the document lifetime, there would be a lot of restyling to do in a
hurry.
Regards,
--
Nicolas Mailhot
