Antonio asked: > On 2003.02.25, 19:36, Asmus Freytag <[EMAIL PROTECTED]> wrote: > > > At 12:55 PM 2/25/03 +0000, Anto'nio Martins-Tuva'lkin wrote: > > > > > Most (all?) of them are composable, either by means of letter + > > > slash (OSLI) or by ZWJ (for things like "Pta" or "Pts", if > > > anything), > > > > Using ZWJ for such things is frowned upon. The ZWJ [is] not a general > > purpose compositor. > > Sorry. I mean such an invisible character that would keep those letters > toghether, even when the inter-character space is expanded, like as if > they were in the same "lead type". (The same thing I'd use decompose > U+0133 into i+THING+j.) > > What Unicode character should be used for this, then? > > > The ZWJ may be used to request a ligature between two characters, > > Isn't this the role of CGJ (combining grapheme joiner)? «Indicates that > the adjoining characters are to be treated as a graphemic unit.»
While the language has been confusing, the intent is the following. ZWJ/ZWNJ are used for control of cursive connection (as for Arabic), to affect exact glyph shaping in various Indic scripts, and to request ligation or non-ligation in various scripts. Think of them as a non-displaying "joining context" which is used by a rendering engine (or font) to influence the exact display of glyphs -- and in particular their visible connection to one another. CGJ (COMBINING GRAPHEME JOINER) is used to connect two (or more) characters together into a *logical* unit for the purposes of some processing. It is intended to create exceptional units (only if required) for processes such as boundary determination or sorting. Think of it as a character "gluer" that has no impact on display, per se. --Ken