Re: [m17n-list] ewts problems

[K. Handa]

In article <address@hidden>, Élie Roux <address@hidden> writes:

> The most annoying bug is explained here:
> https://savannah.nongnu.org/bugs/index.php?42656

I see.

> >  For instance, the change deletes
> > the mapping for characters used for sanskrit,

> It shouldn't be the case, no. There are some simplifications in the
> code, but there should be no such things. Do you have an example of a
> regression?

The current version has a control variable "precomposed".
It is, by default, 1.  So typing "oM" produces a precomposed
character U+0F00 (ༀ).  But if a user customizes that variable
to 0, typing "oM" should produced the character sequence
U+0F6B U+0F7C U+F7E (ཨོཾ).  As your change deletes that
variable, such a customization stops working.

> If you talk about gh which used to produce the same as g+h, this is, in
> my opinion quite wrong, as there is no such thing in the EWTS
> descrption. This is explained in the commits I believe.

Where in your code is it explained?

> > deletes
> > a mapping for U+0F76, etc.

> This character is an equivalent of 0FB2+0F80 (see
> http://www.unicode.org/charts/PDF/U0F00.pdf), and there is no shortcut
> for it in EWTS; the point, as before, is to respect the standard more...

I agree that it is important to respect the standard.  But
keeping backward compatibility is also important.  Are you
sure that there's no need of inputting such characters as
U+0F76?

Here's a reply from Hugues.

> I've downloaded the no mim file. As for the 'brlabs' problem, perhaps  
> there was a mistake in the file, as there is a possibility of wrong typing.

> But the most important in the new file is that it adds numerous  
> possibilities in 'ambiguous cases with xxx as prefix' that are forbidden  
> in tibetan spelling, such as 'grga' 'grnga' ...and so on. So there is NO  
> tibetan word beginning with such prefixed syllables.
> I haven't checked other differences between the last version i gave and  
> this new one.

> As for 'gh' case, which is a case of sanskrit letter, the problem with  
> adding such possibilities (which are not in ewts if i remember well), is  
> that it adds a lot different cases, i.e. buddha -> bud+dha ? buddha ?  
> budd+ha ? bud+d+ha ? an so on with many sanskrit transcriptions. I think  
> that why there is no such possibility in EWTS chart, because it make  
> things simpler, but that's just my opinion.


As I have no knowledge about Tibetan, I can't decide what is
good as Tibetan input method.  

Élie and Hugues, could you two please discuss what is best
for bo-ewts.mim and let me know the conclusion?  If the
current one has a bug, it should be fixed.  If the behavor
of the input method should be customized according to a
user's preference, such a facility should be implemented.
For such customizations, if you explain the detailed spec to
me, I'll implement them.

---
Kenichi Handa
address@hidden