Dov Feldstern wrote:
Helge Hafting wrote:
[...]
Not strange at all. Document text that isn't explicitly set to some
language, is in the "document language" and changes if that
language is ever changed. This is very convenient.
I can see that this will be different with languages that uses
different writing systems, such as english and hebrew. Changing one
to another might be meaningless with no common letters.
But then, anyone wanting to type hebrew will notice right away that
their new document is set to english.
I guess this is the crux of our disagreement... As you yourself said,
it'll never happen that one types anything in Hebrew and only later
realizes that he's been typing in English, it's just not possible. But
what does often happen is that you start typing a mixture of English
and Hebrew, and then at some point realize that this should really be
a Hebrew document, so you switch the language of the document, but you
certainly don't expect all existing English paragraphs to become Hebrew...
I agree that transforming english to hebrew is wrong.
In that case, I don't dispute the validity of this bug report. But I
do want to clarify that when it is fixed, care should be taken not to
change the behavior regarding RTL languages. I'll add a note to this
effect to the bug report.
I don't think it is RTL is the only special case. We sure don't want to
force
transformations between latin/greek/russian either.
Ideally, I guess the determination of whether or not to preserve the
language of "default" text when switching the document language from A
to B should be something like this: if A and B have the same alphabet
(which I don't think we can check --- but what about if we would use
the encoding as a surrogate --- would that work for you?) then all
"default language" text should remain "default", i.e., it is now
considered to be in language B. However, if B has a different alphabet
(encoding) than A, then all text marked as "default" should now be
explicitly marked as A.
Does that sound right?
Yes. Transformation may be impossible with different encodings - that
is why they are different. So lets not even try in those cases.
Now, utf8 can encode anything, so theoretically, a utf8 document could
be changed from any language to any language. But latex don't
support this today. And if it ever does, this change should actually work
so it can be allowed.
Note, however, that this could be very confusing, because sometimes
the language will change, and other times it won't, and the user may
not understand why it is or isn't in any specific case...
It will always work for some combinations of languages, and never
for other combinations. Few people regularly work with *many*
languages. Those that do, will probably see the sense in that they
can correct "english" to "french" but not turn it into "hebrew" or "greek".
Helge Hafting
