Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: $Revision: 1.6.2.16 $; site ima.UUCP Path: utzoo!linus!philabs!cmcl2!harvard!think!ima!johnl From: johnl@ima.UUCP Newsgroups: net.internat Subject: Re: Escape chars etc. Message-ID: <125100002@ima.UUCP> Date: Thu, 7-Nov-85 22:33:00 EST Article-I.D.: ima.125100002 Posted: Thu Nov 7 22:33:00 1985 Date-Received: Sat, 9-Nov-85 06:31:06 EST References: <485@harvard.UUCP> Lines: 22 Nf-ID: #R:harvard:-48500:ima:125100002:000:1041 Nf-From: ima!johnl Nov 7 22:33:00 1985 /* Written 4:59 pm Nov 7, 1985 by macrakis@harvard in ima:net.internat */ > Representing multiple character sets with shifts has been suggested on > this list. Roy Smith points out that this would cause great > difficulties in using Unix tools, for instance: > > Guess what breaks: > > Tail -- you [have to] track the shift commands. > > Grep -- should grep track the shift commands... > > Do you now need a shift-optimizing filter ...? Sounds to me like it's a bad idea to have shift sequences with global effect, which is not too surprising. If we declare that each line has to be self-describing, with appropriate shift sequences at the beginning if it is in other than the standard character set, that makes it entirely possible to make tail, grep, and such work. I think that we need to examine our conventions and assumptions about our conventions very carefully, and if we do, we can come up with a set that makes it possible to get work done without having to rewrite every bit of code from scratch. John Levine, ima!johnl