- From: Martin Thomson <martin.thomson@gmail.com>
- Date: Mon, 21 Apr 2014 11:30:42 -0700
- To: K.Morgan@iaea.org
- Cc: Matthew Kerwin <matthew@kerwin.net.au>, Patrick McManus <pmcmanus@mozilla.com>, Mark Nottingham <mnot@mnot.net>, C.Brunhuber@iaea.org, HTTP Working Group <ietf-http-wg@w3.org>
On 21 April 2014 11:10, <K.Morgan@iaea.org> wrote: > Thanks for the response. Do all MUST-level requirements have to be enforceable? (If so, I didn't know that.) That's not quite the right question. We use "MUST" to proscribe behaviour so that an endpoint does not do things (ultimately, generate or not generate packets in particular patterns) that might affect a peer endpoint negatively. In general, if compliance with the "MUST" cannot be detected, then there's no point in saying it. Implementations that detect behaviour that contradicts a "MUST" are entitled to do whatever they want with the connection, since it's clear at that point the peer is not talking the protocol they originally thought. The recommendation we make is to send an error message using the original protocol and terminate the connection. For example, there's no way to tell an implementation with a bad server implementation of HTTP/2 prioritization from one with a badly misbehaving back-end. Thus, we don't use MUST when talking about prioritization (or at least, we shouldn't be). So, I think that your answer is "yes".
Received on Monday, 21 April 2014 18:31:09 UTC