Re: H2 vs HPACK header table size... again

On Thu, Dec 23, 2021 at 03:46:31PM +0200, Ilari Liusvaara wrote:
> On Thu, Dec 23, 2021 at 04:21:01PM +1100, Martin Thomson wrote:
> > 
> > Or, you could imagine using that 0.5-RTT data to prime the HPACK
> > header table with useful values (Server, or other common values
> > could be a good use of the bandwidth). 
> Beware that some HTTP/2 clients vomit (IIRC with fatal TLS alert 10
> after sending client Finished) if the server sends 0.5-RTT data (no
> client certificate request). I do not know more specifics, but I
> suspect that at least in some cases, there is some "anti-virus" or
> "security" product involved on the client side.

Yes we've seen that more than once, they used to break WebSocket a
decade ago and more recently to break websocket-over-h2 IIRC. I don't
know what impacts it can have on such mechanisms but we must always
make sure that we can fail fast and in a detectable way.


Received on Thursday, 23 December 2021 14:44:03 UTC