- From: Koen Holtman <koen@win.tue.nl>
- Date: Thu, 20 Nov 1997 22:35:36 +0100 (MET)
- To: Jim Gettys <jg@pa.dec.com>
- Cc: koen@win.tue.nl, masinter@parc.xerox.com, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Jim Gettys: > >Whether at least existing HTTP implementations can possibly use >more than three digits depends strongly on implementation details: > >a) If the implementation takes the code, and then performs a MOD 100 >operation, to get a subcode for the code type, then it makes sense to >allow for more than three digit status codes. > >b) If existing implementations just take the first decimal digit, >and then switch off of the remaining digits, more than 3 digit status >codes are feasible. True, but the draft discusses status codes for `HTTP and HTTP-Derived Protocols', and I suppose that for at least some HTTP-Derived protocols, compatibility with the status code parsers of plain http clients will not be an issue. And if you really want to send a 5 digit code like 45205 to a http client without prior negotiation, you could always invent some wrapping scheme with responses like HTTP/1.1 299 Extended status code Status: 45205 Epibration complete Content-type: text/html .... Koen. > >Without some data on whether implementations (particularly proxies) do >a) or b), any discussion of more than three digit codes is pretty silly. > >I suspect that a) predominates, but having never implemented a proxy >or server, or for that matter, client library, I can't say. > >So implementers, how is the jungle out there??? > > Your editor, > - Jim Gettys > >
Received on Thursday, 20 November 1997 13:39:46 UTC