- From: Martin Duerst <duerst@w3.org>
- Date: Fri, 25 Oct 2002 22:19:05 +0900
- To: Chris Lilley <chris@w3.org>
- Cc: Ville Skytta <ville.skytta@iki.fi>, Terje Bless <link@pobox.com>, W3C Validator <www-validator@w3.org>
At 14:53 02/10/25 +0200, Chris Lilley wrote: >On Friday, October 25, 2002, 2:19:26 PM, Martin wrote: >MD> Sniffing on content and sniffing on extensions are not that different. > >Ah, so you count all server setup as sniffing. That seems tobe a >significant dilution of the term. No, sorry. Any single server is set up the way it is set up. But I thought that what you suggested was that we use the server settings to decide how to type file uploads. While this is okay for actual uploads to a server (because there is no specific protocol for such uploads, and we just have to assume external knowledge), I don't think it's okay for uploads to the validator, which come from any arbitrary third party which may use extensions rather different from those used at W3C. >MD> What I'm saying is that if we want to avoid sniffing, then we should >MD> do that altogether, not replace one way with another way that looks >MD> better for the moment. > >Still not clear - every file on our server is served using what >you call 'sniffing' ... No, there is a difference between server setup based on extensions and sniffing of unrelated third-party files. >MD> The strict ascii default for text/foo+xml >MD> was a rather explicit requirement from the IETF. > >Thus, text/*+xml should never be used because it contradicts the XML >specification. So, it should just be deleted. if they won't delete it, >it should be warned against. I agree that text/*+xml doesn't have that many uses. But it does not contradict the XML specification. For details, please see http://www.w3.org/TR/REC-xml#sec-guessing-with-ext-info. >So as I said, one way to make the validator work best within the >constraints of incorrect MIME types being set by file upload would be >to use our own server mime.types file to create the MIME type, as it >the file was being served from our server (which it is, briefly). Again, if, as it seems, mime types on file upload isn't reliable, I have no problem ignoring them. But pretending that the file is served from our server doesn't look like a solution to me. (The file is never actually served from our server, not ever briefly.) Regards, Martin.
Received on Friday, 25 October 2002 20:05:01 UTC