sorry ... my reply went directly to Matthias Andree instead of mailing list ... fwd. his answer to mailing list as well
David S. -------- Original Message -------- Subject: Re: [Openvpn-devel] [patch] enhance compatibility with HTTP/1.1 proxies List-Post: openvpn-devel@lists.sourceforge.net Date: Wed, 12 Nov 2008 23:41:18 +0100 From: Matthias Andree <matthias.and...@gmx.de> To: David Sommerseth <openvpn.l...@topphemmelig.net> References: <200810271301.m9rd1bkt006...@post.behrens.de> <200811121224.maccomfd077...@post.behrens.de> <20081112133657.ga15...@merlin.emma.line.org> <491b56d4.9030...@topphemmelig.net> On Wed, 12 Nov 2008, David Sommerseth wrote: > Completely agree ... this seems to be a more sensible way to approach this > issue. Seems to be > pretty fail-safe too. In this approach it really only depends on the > contents of > p->options.http_version and how sscanf(...) parses it. And it should be fair > to believe that > sscanf(...) will do a pretty good job. > > Should maybe do some testing how this snippet will work with both sensible > http versions (0.8, 0.9, > 1.0, 1.1, 1.2, 1.00, 1.10, 1.01, 2.0), but also with real garbage (1.1a, > GARBAGE, 1.) Expected output: 1.1a will be parsed as 1.1, GARBAGE and 1. will be rejected because rc is 0 or 1, respectively. Otherwise, 0.8, 0.9, and 2.0 should be rejected; 1.0 and 1.00 should use the 1.0 code branch; 1.1, 1.2, 1.10, 1.01 should use the >= 1.1 branch on the assumption that >= 1.1 aren't changing Host: semantics. Let me know if there are differences. -- Matthias Andree
