Re: [m17n-list] ewts problems

[Élie Roux]

> But, again, one thing is to say "this is valid Tibetan" and another to
> say "this is valid EWTS". I don't think EWTS should be sacralized as
> truely representing the essence of Tibetan, but more a convenient way to
> type Tibetan letters. For instance, suppose there was the opposite, a
> kind of EWTS for Tibetans to typeset French letters with Tibetan
> characters. In french "inpo" (as in Rinpoché) is invalid, as there is a
> grammar rule saying that before a "p", "n" is transformed into "m"...
> Well, I don't believe this kind of rule should apply to a French EWTS,
> it just complicates for nothing... Do you see my point?

Well, I'd like to push the reasoning a bit further, as I already had
this discussion with the people at the origin of EWTS when I did my
first EWTS parsing library (https://github.com/eroux/ewts):

EWTS (Wylie actually, but the distinction here is not really relevant)
was first invented to allow westerners to easily write in Latin
characters the informations contained in Tibetan script. So it's an
output, not an input method. Ironically, EWTS is so good that it conveys
even more information than Tibetan script itself! For instance mangs and
mngas, two different and valid Tibetan words (at different places in a
dictionary) are written the same in Tibetan (མངས་), while EWTS
differentiates them! EWTS fully filled this function for decades, but
with the spread of Unicode (and, most importantly software support for
hindic scripts in OTF fonts), there is no need for EWTS in this sense
anymore, there are Tibetan keyboards in smart-phones and tablets, and
people are working in Unicode script directly, no need for EWTS.

But EWTS is still used by westerners like us who want to type Tibetan on
a Western keyboard (Tibetans use a different keyboard). So there is no
need for it to convey the very precise informations of Tibetan script,
needless to say that there is no need to have more information that in
Tibetan script! The goal is for it to be an intuitive input method.
Sadly, as it was not thought this way, there are lacks in the
specification (provoking our discussion). I asked people at thdl to fill
these lacks, but they're busy and don't have time for this...

Having stated this, let's take the case of "drda":

 * first, in the continuation of the previous discussion, you're right,
this is absolutely not Tibetan. So a Tibetologist noting དརྡ་ in wylie
would use darda, to show that this is sanskrit, and not drda, because
drda would confuse other people used to read Tibetan in EWTS. But, here,
we don't use EWTS as the output, but as the input, so, do we care about
drda being a bit confusing for people used to read EWTS? I don't think
so, as long as we also support darda.

 * second, what does the standard say about drda? If we follow the rules at
http://www.thlib.org/reference/transliteration/#!essay=/thl/ewts/rules/
well, there is absolutely nothing... EWTS was thought as an output and
not as an input, so it doesn't handle this kind of ambiguous cases...

 * third, if the user types drda, what does he expect? Hard to guess...

Well, I didn't expect that, but it seems I countered your argument, but
I countered mine too...

So would that be ok for me to send another patch removing all these
non-Tibetan cases?

By the way, maybe a set of git patches is more difficult for you to
handle than the file directly... Would you like me to send the final
file directly?

Thank you,
-- 
Elie