Re: HTTP/2 and Websockets

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 25/11/2014 9:16 p.m., Yutaka Hirano wrote:
> Hmm, OK, I agree that "magic bad proxy" is a magic word.
> 
> Let me confirm: 1 We use http2 DATA frames to convey WebSocket
> data. 1.1 Each http2 frame payload consists of one-byte WebSocket
> header and WebSocket data payload. 2. All intermediaries MUST
> understand WebSocket (i.e. If there is an intermediary that doesn't
> understand WebSocket, the opening handshake must fail).
> 
> Do you agree with these?
> 

(2) "all intermediaries" does not seem possible. To achieve it an
intermediary receiving a SETTINGS negotiating WS extansion must
already know the destination servers and upstream contact it will have
for all possible WS requests/connections sent by that client.


IMHO (2) should be phrased along the lines of intermediaries
advertising WS extension to HTTP/2 MUST be capable of fallback to WS
over TCP (or other such as HTTP/1.1) if any of their upstream peers do
not negotiate WS extension as well.

That way if anyone offers WS support they are guaranteeing that the WS
connections sent to them will be possible one way or another. It is
not the end clients business how they achieve WS, so long as they do.

Amos
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (MingW32)

iQEcBAEBAgAGBQJUdFxtAAoJELJo5wb/XPRjjSEIALUD+ZERvbKsKqsN3XZqZ+Fn
ogFDANbTeKcMjAH/0FNKLDF2+SlSeGFIvMT+cVtUBg7Cr8JSglPYpFea/ylP/K1n
rKzDoN+nrM8zvLTf10+sjirlojkH2gtKc+QDKKDJnYAFyQhw75u2DiN5b31YmqZ+
ZU9pQB8eyMF9vBVTtsle7wP4gujub8kXIDgDZtgGUnxYiAi1pAJlKIzMI3lfrogd
+QMLt5nxZZe8EMUcLiUe6NC8xfORGQkHPq4rvDnz6Zz+62k1wkzjzzsoJfbJDspr
9J0cPleAzwTjryz80HUSca2o4amSTGdxeux+hB79szD0BxA6fRBRxDrEjf8hRcU=
=0xoT
-----END PGP SIGNATURE-----

Received on Tuesday, 25 November 2014 10:40:33 UTC