- From: Olivier Thereaux <ot@w3.org>
- Date: Thu, 9 Oct 2003 15:12:27 +0900
- To: Ilya Konstantinov <www-validator@future.shiny.co.il>
- Cc: www-validator@w3.org
Ilya, On Tuesday, Oct 7, 2003, at 23:56 Asia/Tokyo, Ilya Konstantinov wrote: > Olivier wanted to continue this discussion in the list. The following > message is a clarification of my recent bug report. Thanks for following this up on the mailing-list. It helped me in discussing with others (Thanks Terje and Nick) and finding out what actually happens. >> 'check' will be happy to validate whatever I throw at it, either HTML >> or CSS, according to the Content-Type. To be precise, it won't >> validate >> itself but redirect me to the HTML validator or the CSS >> validator. So far so good... >> BUT - what if there's both CSS and HTML at the given URL, and my >> server >> determines which type to send you via content negotiation (a.k.a. >> Accept-Content header)? Then 'check' will have to express a >> preference towards either 'text/html' or 'text/css'. >> >> FACT - 'check' seems to use the CSS Validator's code for fetching my >> page, and the CSS Validator prefers 'text/css' over 'text/html'. Not quite. The fact is that the Markup validator does not "prefer anything", it just redirects to the CSS validator (which does prefer CSS) when given a CSS resource to validate. The problem here comes more from your side as it does from ours: - you are negotiating, on your server, CSS and HTML resources. Negotiating should be done among equivalent resources in different formats or languages, which means negotiating some content and their stylesheet is a bad (abusive) use of format negotiation. - The markup validator does not (currently) express any preference. As far as I can tell it doesn't send any Accept: header at all (need to check, though) - Given the above, your server drives the negotiation and seems to prefer CSS to HTML. - The negotiation gives CSS, and the Markup validator redirects to the CSS validator, which does the rest (send the user-agent string and accept header your recorded, etc.) I do acknowledge that to cope with such (wrong) situations the Markup Validator should probably send a Accept: header stating a preference for content it can validate itself, but the situation should, generally, not happen if no-one abuses format negotiation... -- olivier
Received on Thursday, 9 October 2003 02:12:27 UTC