From: Antoine Leca (Antoine10646@leca-marti.org)
Date: Thu Jun 02 2005 - 10:30:17 CDT
On Thursday, June 2nd, 2005 14:05Z Philippe Verdy wrote:
> From: "Antoine Leca" <Antoine10646@leca-marti.org>
>> On Wednesday, June 1st, 2005 12:21Z Dominikus Scherkl wrote:
>>>>> - - ligature processing is a required feature to support
>>>>> even legacy ISO 8859 charsets like Arabic, or Indian
>>>>> standard charsets (ISCII).
> Yes I know, and I did not use the term "script" but "Indian standard
> charsets". This is really different...
Oh, fine point I missed.
Yet, I do not see the reason for the plural. As far as I know, there is only
one "Indian standard charset", which is ISCII-91. And the only other one
(beyond Unicode, that is) that might applies for the plural, namely ISSCII,
was clearly superceeded and anyway is pretty out of use nowadays (OTOH, this
latter ISSCII really requires ligatures, since it only deals with Nagari.)
> A "charset" is definitely not a script, but an encoded subset of
> abstract characters, possibly belonging to one or more scripts, and
> serialized on a plain-text stream with a predefinite order and
> predefined values of bytes.
Sort of agreed. Such as, if we consider ISCII, -m̐/-˜, -ṁ, -ḥ, a, ā, i,...
ai, ka, kha, ga, gha, ṅa, ..., -a, -ā,...
And then, it is quite required some mechanism, for example to display ki
when the string is <ka, -i>.
But it is not a ligature, at least not what people here expects when
speaking about ligatures.
(OK, ISCII also defines various charset-encoding's, just like Unicode
defines UTF-16, UTF-8 etc. I did not consider this, since I do not believe
this adds clarity.)
> And I did not imply that Latin characters were needed in those Indian
> ISCII charsets,
Sorry, I cannot parse your point.
The above is perfect ISCII, yet it is only Latin characters; you can replace
the glyphs with the one for other script such as Devanagari (and then the
rules to render will differ, here for example the order of the glyphs will
be reversed, a bit like if I wrote "ik").
> just that most of these charsets (as well as Arabic)
What is the reason for the plural here?
Again, ISCII-91 is just ONE charset, covering many scripts.
> already need a ligature processing system,
OTOH, ISCII needs _various_ processing (rendering) systems, about one per
script; yet, those for Roman and in a limited way Tamil does not precisely
Also, those ligature systems are context dependent (something that is not
very common), in the case of the letter ra (BF) _very_ context dependent,
which calls for pretty ad hoc mechanisms (such as the rearrangements in GX,
or worse the Indic OpenType specification*s*, which are barely transposable
even between "close" scripts.)
> So the technology is already there, and it seems unreasonable to not
> use it when rendering Latin,
That is another point.
Nowadays, if you try to render Indic scripts with advanced typographical
tools (or even sometimes not-so-advanced ones), you are pretty barebone. For
example, if using X11 you are required to use various hacks to achieve the
most basic rendering (I know this is changing fast ;-).) Even if the
underlying hardwares make this pretty moot, this implies a significant cost
at the performance level.
OTOH, if you are doing Latin/Greek/Cyrillic typography or pre-press, the
panorama is completely different, not one tool is common, you have plently
of parameters to play with, some of which would be irrelevant for Indic
languages (think about weight, change of casing, slanting, kerning, baseline
adjustment to name a few). Yet it is this very market which is dealing with
ligatures in the Latin script, "normal" users do not care.
So, sorry but no, the technology is not really there.
And now, you can turn to Avallon which _aims_ at making such a technology
available, at least that's Paul's stated objective.
In a few years from now.
This archive was generated by hypermail 2.1.5 : Thu Jun 02 2005 - 10:31:13 CDT