- From: Etan Wexler <ewexler@stickdog.com>
- Date: Thu, 19 Aug 2004 22:12:46 -0700
- To: www-style@w3.org
Justin Wood wrote to <mailto:www-style@w3.org> on 17 August 2004 in "Re: [CSS21] Font-family syntax" (<mid:4121C2EE.1060608@bacon.qcc.mass.edu>): > Etan Wexler wrote: > >> There are many syntactic constructs which, though allowed by the >> prose description of 'font-family', are forbidden by the Appendix G >> grammar ... > > Does not a quoted string solve all these issues? No. Using <string> lexical types makes things convenient for style-sheet authors and for authoring tools. The issue remains outstanding for implementors of CSS parsers and for the specification's editors. Nothing about strings tells people whether the following is a valid style sheet. example { font-family: Trouble: A (Very) Troublesome Type Family & Interesting Prospect. ; } It's the prose and the formal grammar that tell people whether the preceding is a valid style sheet. Unfortunately, there is contradiction between the two. Mandating the use of the <string> type for all font-family names since the solidification of CSS1 would have avoided this problem, sure. But the decision to allow bare names was made circa 1995/1996, was made with a good share of reason, and is extremely unlikely to undergo reversal now. In searching for the original discussion over the syntax of font-family names (which I failed to find), I found something that Bert Bos wrote in one of his many moments of prescience [1]: "[We] should be very careful that we don't make any mistakes, because we probably can't fix them later." [1] Bert Bos. "The style agenda". 30 May 1995. Public correspondence on <mailto:www-style@w3.org>. <mid:199505301915.AA262011341@freya.let.rug.nl>, <http://lists.w3.org/Archives/Public/www-style/1995May/0010.html>. -- Etan Wexler.
Received on Friday, 20 August 2004 05:15:01 UTC