From erik@sran8.sra.co.jp Tue Feb 19 06:54:44 1991 Received: from mcsun.EU.net by dkuug.dk via EUnet with SMTP (5.64+/8+bit/IDA-1.2.8) id AA24326; Tue, 19 Feb 91 06:54:44 +0100 Received: from [133.137.4.3] by mcsun.EU.net with SMTP; id AA01860 (5.65a/CWI-2.74); Tue, 19 Feb 91 06:57:39 +0100 Received: from srava.sra.co.jp by srawgw.sra.co.jp (5.64WH/1.4) id AA16640; Tue, 19 Feb 91 14:59:03 +0900 Received: from sran8.sra.co.jp by srava.sra.co.jp (5.64b/6.4J.6-BJW) id AA15768; Tue, 19 Feb 91 14:57:03 +0900 Received: from localhost by sran8.sra.co.jp (4.0/6.4J.6-SJ) id AA10461; Tue, 19 Feb 91 14:55:38 JST Return-Path: Message-Id: <9102190555.AA10461@sran8.sra.co.jp> Reply-To: erik@sra.co.jp From: Erik M. van der Poel To: i18n@dkuug.dk Subject: Re: (i18n 74) Re: paper presented to WG11 (#20) Date: Tue, 19 Feb 91 14:55:36 +0900 Sender: erik@sran8.sra.co.jp X-Charset: ASCII X-Char-Esc: 29 > char *str = "s\ndag"; > [...] > 3. Compile the \< token into an otherwise unused character > code and use that at run-time to trigger "on the fly" > substitution in all STANDARD routines operating on strings, > e.g. strcpy and printf. > > Kim F. Storm We should try to resist the temptation of attaching special meanings to "otherwise unused" codes. One of the versions of the cpp program used the code 0xfe for something special. This caused core dumps when used with Japanese characters. (Why have *another* special code? Why not just use `\'?) [Of course, if the combination `\<' causes a core dump, we're still up the creek...] It is of course easy to modify e.g. strcpy and printf to handle the new escape sequences, but frequently the programmer needs to process strings within his/her own program, rather than within a "standard" routine. Simple operations like counting characters become rather complex when escape sequences are present. So, in this sense, it would be easier to use a fixed-width, stateless code like 10646's 4-byte form. E.g.: wchar_t *ws = L"s\ndag"; Even when the programmer insists on using `char' instead of `wchar_t', we need to make sure that strings read in from a file and strings compiled into a program are encoded consistently. This could be done by having a compiler that compiles strings into, say, 10646, and having files in 10646 too. Or we could leave the strings as they are and store files in a -like format, but I don't think this is Keld's intention. Of course, interpreting the strings at compile-time will hard-code info from a particular charmap, but then that is one of the reasons for having a universal coded character set. :-) Erik M. van der Poel erik@sra.co.jp Software Research Associates, Inc., Tokyo, Japan TEL +81-3-3234-2692