From ALB@immedia.ca Sat Sep 20 08:49:00 1994 Received: from Clouso.CRIM.CA by dkuug.dk with SMTP id AA25160 (5.65c8/IDA-1.4.4j for ); Tue, 20 Sep 1994 15:55:38 +0200 Received: from immedia.ca by clouso.crim.ca (4.1/SMI-4.1) id AA28622; Tue, 20 Sep 94 09:55:32 EDT Return-Path: Received: by immedia.ca (3.2/2.D) id AA27961; 20 Sep 94 13:51:28 -0500 Date: 20 Sep 94 13:49:00 -0500 From: ALB@immedia.ca Message-Id: <199409201351.AA27961@immedia.ca> To: i18n@dkuug.dk Cc: cpwg-mail@revcan.ca Subject: About the relativity of character identification X-Charset: ASCII X-Char-Esc: 29 ---------- Copy of a note I just sent to ISO10646 list server, important in my opinion. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ >The point is that G WITH CEDILLA is _not_ a valid or descriptive >name for G MI/KSTINA/JUMS. G WITH COMMA BELOW is. > >>If multiple names were >>allowed then adding a new one would not break any existing code. > >I wholeheartedly agree that a registry of aliases should exist. I could not agree more. I am revising a Canadian standard now which must use a stable identifier for characters. Because of this business of LETTER or LIGATURE or JOINED DIGRAPH AE (and perhaps others, what about OE?) which is not fixed yet, we have decided to use the Uxxxx[xxxx] (where xx is a hexadecimal digit representing the code point of ISO/IEC 10646 used as a catalog number to preserve coding independence) identifiers to relate the names we use in both languages (French and English) to the right characters in the UCS. Furthermore as we make our standards in 2 languages, we have at least already 2 sets of names for each character and we will not publish double-size documents (think about the discussion on ISO/IEC 10646 price) just because only one set is official (which it has no reasonable right to be). Of course in a single language in ISO standards names must relate between different standards (the names of the English version of ISO/IEC 10646 must fit with the names of, say, the English version of ISO/IEC 8859 [all parts], and the names of the French version of ISO/IEC 10646 when published must fit with the names of, say, the current or righted French versions of ISO/IEC 8859). Names remain the best anchor point between different standards, provided the absolute of this remains inside the documents themselves tied to the language in use. For the outside world the anchor point should be the ISO/IEC 10646 "catalog numbers" which are constituted by their UCS coding. Two other projects of international standards that I know of will use Uxxxx[xxxx] identifiers for another reason: the names are simply too long and often won't fit on an Internet screen line (this creates syntax problems everywhere and is in practice [for all those I know who have tried] not universally applicable). Keld's system of short identifiers (provided there are also aliases, which Keld agreed about in principle) is also of help to complement Uxxxx[xxxx] scheme (but instead of standardizing the short names, we use them as names of variables whose definition uses Uxxxx[xxxx])). Perhaps we could use Uxxxx[xxxx] all the time to make sure different naming or identification schemes all fit together. This would also help solving interlanguage and intercultural problems like the one of Latvian-view names of characters in English versus Turkish-view names also in English. Also in Canada we have problems with names of natives' syllabic (unified) characters: in many native languages, the names for a same character are quite different and we can't even use the transliterated values as this would mean many transliterations for one character: if we have to do this there shall be aliases, no question about it. Alain LaBont Secrtariat du Conseil du trsor Gouvernement du Qubec