- From: Al Gilman <asgilman@access.digex.net>
- Date: Thu, 10 Jul 1997 12:16:02 -0400 (EDT)
- To: w3c-wai-wg@w3.org (WAI Working Group)
I believe that the HTTP HEAD request only get you the RFC-822-ish HTTP message head, not the HTML <HEAD>...</HEAD> block. On the other hand, Brian Behlendorf did coin the following tempting dodge: ----- Forwarded message from Brian Behlendorf ----- Date: Sun, 11 May 1997 17:25:44 -0700 (PDT) From: Brian Behlendorf <brian@hyperreal.com> [...] Another response I'd give is that if I was writing the client, one thing I could do would be to use the "byterange" part of the HTTP/1.1 spec and only access the first, say 500 bytes of a document or object, on the premise that any <TITLE> is likely to be there 90% of the time. So long as all the HTTP headers and the first chunk of text can fit inside 1000 bytes, then it'll just be one TCP/IP packet; the majority of delay when connections start up come from waiting for a second or third packet, it doesn't make much difference if the first packet is 5 bytes or 1000. But that's a nit I guess. On a larger scale, the question is, is there a framework for general metainformation about internet objects. Some people are starting to make noise that PICS could be used as a framework: http://www.w3.org/PICS/ ------------------------------------------------------------------------ By sending a GET with a BYTERANGE restriction one could get a single- packet sample, and test for <HEAD> closure in the fragment returned (similarly for PNG if the description is in the head of the file). One could iterate in cases where the GET of the <HEAD> failed to be complete. That would get all present <TITLE>, <META> and <LINK> information. So not all the meta-data based methods require new filing conventions or the registration of MIME headers. -- Al Gilman
Received on Thursday, 10 July 1997 12:16:06 UTC