- From: Gary Adams - Sun Microsystems Labs BOS <gra@zeppo.East.Sun.COM>
- Date: Thu, 27 Feb 1997 14:24:43 -0500 (EST)
- To: www-international@w3.org, misha.wolf@reuters.com
# From: Misha Wolf <misha.wolf@reuters.com> # Subject: Re: Language label # # For a reason I haven't yet understood, Tomas considers that the server should # take this information from inside the HTML document. Others, including me, # suggest that this information belongs outside the document, eg in the file # name: # # home.fr.html # home.de.html # home.zh-cn.html # # That is what has to be done for other files, such as images, eg: # # go-home.fr.gif # go-home.de.gif # go-home.zh-cn.gif # # I don't understand why HTML files should be treated differently. This is a recurring discussion about what should be "in" the document, what should be the responsibility of the "container" which holds the document or what is the responsibility of the "transport agent" that delivers the resource. I wish I could point to the definitive document that deals with these issues in greater detail, but it has gone on several times in the "object" forums and in the UR* mailing lists. The best argument for <HTML LANG=xx> is that the document is self describing and can be transported via http, ftp, mail, gopher, webnfs, imap, etc. and still have proper identification in the web client application (could be a browser, could be indexing engine, could be a translation service, etc.) The best argument for <META HTTP-EQUIV...> is that document authors can add information to the http headers without special actions taken by a site administrator. This is at best a stop gap remedy for adhoc transmission of http documents. The larger problem has to do with how "features" in general are transmitted about doucments out of band with the "name of the resource". Imagine foo.fr.tables.color.frames.html or worse encodings of a particular document feature set in the name of a file. (See http://www.acl.lanl.gov/HTML_WG/html-wg-95q1.messages/0620.html the Mar 1995 thread about deploying tables to get an idea how ugly adhoc feature labeling can get). Some http servers are already providing simple mechanisms for including metadata in the request, "just outside" of the document itself. e.g. The Apache server implements a "variant file" and an "as is" mechanism to deliver specific http headers or a negotiated match for a particular http request: http://www.apache.org/docs/mod/mod_asis.html http://www.apache.org/docs/mod/mod_negotiation.html There are similar mechanisms in the w3c sample httpd and the Jigsaw server negotiated content filter. (The commercial servers may be lagging behind the freely available daemons at this time.)
Received on Thursday, 27 February 1997 14:29:13 UTC