- From: Amos Jeffries <squid3@treenet.co.nz>
- Date: Sat, 5 Aug 2017 20:32:46 +1200
- To: Mark Nottingham <mnot@mnot.net>
- Cc: ietf-http-wg@w3.org
On 05/08/17 15:55, Mark Nottingham wrote: > HTCPCP was a joke protocol to point out all of the bad ways that > people try to reuse HTTP, so there's a certain amount of irony here. > I don't think a couple of deployments of a joke is enough to justify > consuming a code point in such a constrained name space. > > IIRC we discussed this as part of BIS and explicitly decided not to > reserve 418. > > Now, we could make an argument to skip over it now and use it when > we've exhausted other 4NN code points, but personally my inclination > is to do it now; if we don't want it to ossify, the earlier the > better. It seems to me that it would be best for any formal definition of 418 at least be consistent with the well-known meaning from the joke - then people running across the wrong RFC for it will not be too confused. There are many non-joke things that are rejected already; 406, 414, 426, 501, 505 etc. so something serious rejecting the clients choice of protocol is not exactly out of place. "418 Protocol Unsupported" with a definition between 426 (but not requiring Upgrade) and 505 (but covering the whole choice of protocol, not just its version) would be of some benefit to most middleware and servers I imagine. For example Squid cannot use 426 due to Upgrade, but we still have to deal with multiple weird protocols arriving on port 80 and 443. Today we 400 them simply for lack of anything more specific and helpful. Amos
Received on Saturday, 5 August 2017 08:33:13 UTC