- From: Larry Masinter <masinter@parc.xerox.com>
- Date: Tue, 2 Jul 1996 23:26:00 PDT
- To: erik@netscape.com
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
> This is great, but I get this feeling that we need to "push" people a > bit more. There are a lot of SHOULDs and MAYs in your prose, but no > MUSTs. How about having some MUSTs for the receiving side of things. > I.e. HTTP/1.1 clients MUST be able to accept explicit charset > parameters. Same for HTTP/1.1 servers (and their CGIs). It may be a good > idea to list some of the common cases, e.g. "text/*" subtypes, > application/x-www-form-urlencoded, maybe even multipart/form-data? I thought about how such a requirement could be worded, but it was difficult. First, not all subtypes of text/ take charset parameters. It's actually up to the individal media type. So we could make a rule for text/plain and text/html, but leave the rest alone. But what is it that clients MUST do with text/html;charset=ISO-8859-1? Display such characters 'properly'? Offer to save such files to disk? It got hard to figure out what it was we were going to require. application/x-www-form-urlencoded doesn't have a "charset" parameter, although if something were going to be registered, it 'should', and multipart/form-data just says that the content consists of things with internet media. Anyway, I'd rather just leave it alone. We've at least made it that senders SHOULD send charset parameters, and recipients, well, they better deal with what senders SHOULD send. It was wishful thinking to hope that someone would go and red pencil all of the SHOULDs and MUSTs in the current draft, but.. I wish ... Larry
Received on Tuesday, 2 July 1996 23:29:17 UTC