- From: Willy Tarreau <w@1wt.eu>
- Date: Wed, 19 Oct 2016 16:45:43 +0200
- To: Matt Menke <mmenke@google.com>
- Cc: Frederik Braun <fbraun@mozilla.com>, Daniel Stenberg <daniel@haxx.se>, Poul-Henning Kamp <phk@phk.freebsd.dk>, HTTP working group mailing list <ietf-http-wg@w3.org>
Hi Matt, On Wed, Oct 19, 2016 at 10:30:30AM -0400, Matt Menke wrote: > With Chrome, about 0.003% of main frame responses are detected as HTTP/0.9, > and 0.005% of subresources are HTTP/0.9. Since HTTP/0.9 includes no > explicit identifying headers, some of these could be broken servers > responding with nonsense data. > > We actually did try to go it alone in remove HTTP/0.9 support, but soon > backed off after trying it in our pre-release channels, because of the > aforementioned incident with the configuration page of a line of home > routers. That's very useful data. On haproxy, we never ever had a single report of a broken application caused by the fact that we block 0.9 responses. So your observation of home routers using it tends to confirm our observation in that it is probably only found in cheap consumer devices (that we never see) and never on the hosting side (where haproxy is normally installed). It could be useful to know if Squid sees such responses, because in general it will not be used to access local network devices but will still face the internet and a wide spectrum of self-hosted servers that we don't see. Regards, Willy
Received on Wednesday, 19 October 2016 14:46:43 UTC