From ALB@immedia.ca Fri Feb 25 12:32:00 1994 Received: from Clouso.CRIM.CA by dkuug.dk with SMTP id AA04531 (5.65c8/IDA-1.4.4j); Fri, 25 Feb 1994 18:34:20 +0100 Received: from immedia.ca ([192.139.197.1]) by clouso.crim.ca (4.1/SMI-4.1) id AA29912; Fri, 25 Feb 94 12:33:18 EST Return-Path: Received: by immedia.ca (3.2/2.D) id AA22841; 25 Feb 94 17:39:39 -0500 Date: 25 Feb 94 17:32:00 -0500 From: ALB@immedia.ca Message-Id: <199402251739.AA22841@immedia.ca> To: i18n@dkuug.dk, iso10646@jhuvm.hcf.jhu.edu, rtfq-forum@crim.ca, sc22wg20@dkuug.dk, tc304@dkuug.dk Cc: bealle@torolab6.vnet.ibm.com, cpwg-mail@revcan.rct.ca, umavs@torolab6.vnet.ibm.com Subject: Multiscript Ordering -- Survey on a fundamental issue X-Charset: ASCII X-Char-Esc: 29 ---------- I would like to have as many people state their preference on the following: 1) Is it acceptable, in an eventual international standard on multiscript ordering, to constrain applications to separate different scripts (ex.: Latin, Greek, Cyrillic, Han, Thai, Arabic, Hebrew, and so on) into different fields to be compared, so that only one script set would be processed at a time? 2) Is it necessary to have no restrictions at all and that any given field contain characters from different scripts, given that, anyway, script tables will not be intermixed? (nuance: if one wishes, different scripts could be intermixed, and considered one script for a given application, with the envisaged tailorability feature; this should be considered a special case of single script, even if artificial, with a single set of processing attributes). The reason for this question is that different scripts might have different processing attributes for ordering which should not normally be mixed; if scripts are not mixed in a given field to be compared, this does not pause big technical difficulties and the POSIX model can be applied (except that for characters formed by combining characters, this will require a mod of POSIX specs, unless the full list of normal combinations is known, which is not the case so far). If, on the other hand, scripts are mixed, and if one wants all processing attributes to be taken into account for each script, then POSIX specs will have to be changed or they won't be upward compatible with the multiscript ordering model in preparation. There are solutions to both preferences (I already presented my views in the 4th UNICODE implementation workshop, end of 1992, in Frankfurt), but there are compatibility impacts in the second case more than in the first. The first preference would allow far more efficient implementations too, but that should not be an absolute criteria to preclude option 2 if this is a strong requirement. I wish I am not too obscure to be understood by all the forums to which I am sending this request. Alain LaBont Gouvernement du Qubec Secrtariat du Conseil du trsor 1500B, boul. Charest Ouest, 1er tage Ste-Foy (Qubec), Canada G1N 2E5 Fax: +1 418 646 3571 Tel: +1 418 644 1835 Email: either ALB@IMMEDIA.CA or ALB@SHE.ORG.UK Editor of Default Tailorable Ordering Standard for the Universal Coded Character Set (so called UCS, i.e. ISO/IEC 10646) (Ordering Project: ISO/IEC Project 22.30.02.02)