Peter Kirkwrote:I considered it but now I think it is not the best way to go. Ask others for their views.
Kent Karlsson wrote:
No, I think ZWJ may be exactly the way to go here.
<consonant, holam, (accent), ZWJ, alef/vav> for making a 'ligature'
(of sorts, in a technical sence) where the holam is displayed on the
alef or vav. Without the ZWJ, the holam would be displayed on the
<consonant> to which it logically belongs. (alef and vav are base
characters, so the ZWJ is not breaking any combining sequence here.)
This is an interesting idea, but I don't think it quite works. The decision whether to shift holams on to a following alef or vav is not something to be decided on a per occurrence basis
Then why do you and others propose to add a new character?
There are two possible scenarios with different mechanisms. One mechanism has been shown to work, the other may fail:
but enforced on all renderers. Rather, it is a rendering decision which should be applied consistently, depending on context, through a whole document or style. So it is not something to be encoded, but something which should depend on the font etc.
But you are saying that current mechanisms that try to be that
"smart" sometimes fail.
1) Encode holam male as holam - vav. The algorithm to shift holam to a following vav with no other vowel works and has already been implemented in Ezra SIL (though I di find a small but fixable bug). A special mechanism can be used to force the holam male rendering in a context where it would not be regular, cf. use of ZWJ and ZWNJ to force select particular Arabic forms. But this would never be necessary in normal Hebrew text. So no problem here.
2) Encode holam male as vav - holam. The algorithm to disambiguate this one from vav consonant with holam haser is complex and ambiguous. I suppose ZWJ, ZWNJ, CGJ etc could be used to disambiguate if necessary, but this disambiguation is necessary in regular text and not just in special isolated forms, so the whole thing gets a bit messy. In fact the best way to disambiguate in this case would be to encode every holam male as vav - ZWJ - holam or whatever. But then I see below that you don't seem to allow any of these characters in such a case.
But then you don't need the ZWJ as vav, holam is already distinct from holam, vav. Or does it serve some purpose which I haven't grasped? Remember that I don't want an encoding distinction from the holam being above the consonant, as this is a rendering issue.
The distinction which does need to be encoded is between holam male and vav followed by holam,
Which my proposal trivially acheives:
vav-holam: <vav, holam>
some consonant followed by holam male: <[consonant], holam, ZWJ, vav>
Understood.
as many (though not all) renderers want to make this distinction but they can only do so if there is an encoding difference. It would certainly be an option to encode holam male as vav - ZWJ - holam
This breaks the combining sequence, and is therefore a non-option.
True. That's why I prefer the encoding numbered 1 above. But others insist on this order on the grounds that the mark must come after the base character with which it is graphically associated. As two principles conflict, one or other must be given precedence.
or vav - CGJ - holam.
This breaks the logical order of characters, ...
... and (mis)uses the CGJWell, if you don't like to use CGJ and want to deprecate it, perhaps we should define a new character which has the right properties and intentions to go here. I would be interested to see in precisely what ways such a character would differ from CGJ.
(which would be better deprecated, despite the desparate attempts
at making it useful).
This sounds good. Perhaps we could have the same sequence with ZWJ for the holam to be on the vav even when the vav is vowelised, which is required for the version of the divine name which I posted.
But if this holam is to be encoded before the vav, the ZWJ is redundant.
It seems to be more of a spelling difference than a font difference.
But one could have the convention that <[consonant], holam, (accent),
vav>
always got the holam on the vav, even if the vav is vowelised; and
use <[consonant], holam, (accent), ZWNJ, vav> to put the holam on
the [consonant], regardless of if the vav is vowelised or not.
And then if we want a word initial holam male e.g. for a list of glyphs or just possibly for some other language, I suppose we would have to encode it something like ZWSP - holam - ZWJ - vav, or use WJ (2060 word joiner) instead of ZWSP. Or do you have a better suggestion for this?
I now think I agree with you, Kent, at least on the principle if not on the detail. For a few hours I was on the new character bandwagon, but now I have dropped off it. But let's see what others have to say.I see absolutely no reason whatsoever to encode any new character for these cases. (The cases with two vowels on a Hebrew consonant is quite different.) But you have to choose if the ZWJ solution is the best or the ZWNJ solution is better.
/kent k
PS Why does everyone jump on the option here to encode a new character? It is quite clear that none is needed for this! But everyone is appalled at the prospect of encoding new characters for solving the double vowels problem (where encoding new characters in a smart way appears to be the very best solution). Strange...
-- Peter Kirk [EMAIL PROTECTED] http://web.onetel.net.uk/~peterkirk/
