- From: Martin Thomson <martin.thomson@gmail.com>
- Date: Wed, 27 Aug 2014 11:34:28 -0700
- To: Lucas Pardue <Lucas.Pardue@bbc.co.uk>
- Cc: "ietf-http-wg@w3.org" <ietf-http-wg@w3.org>
On 27 August 2014 10:29, Lucas Pardue <Lucas.Pardue@bbc.co.uk> wrote: > I understand the connection preface has been crafted in order to provide > robust handling of several use cases; such that a range of response codes > may be valid dependent upon situation, however may it be useful to expand > the paragraph in Section 3.5 to capture this succinctly? Firstly, this is only an issue with cleartext. ALPN is hard to get wrong, and that's required when you use TLS. The preface is crafted to ensure that servers drop the connection, or at least fail spectacularly, rather than break in subtle ways. If you are building a server with code that has specific knowledge of HTTP/2, it's probably better to do a little more work to send the connection preface yourself and then presumably immediately send whatever we do to resolve https://github.com/http2/http2-spec/issues/496 [fallback to HTTP/1.1] Or, you could choose to fail spectacularly and hopefully trigger some fallback logic.
Received on Wednesday, 27 August 2014 18:34:57 UTC