ok, I lol'd at that, but I'm willing to admit that the combination is
silly, and that implementations that my proxies are likely to encounter
will get it right in the vast majority of the time.
However, I imagine that many folks would get this wrong if they don't pay
attention to it specifically.
-=R
On Thu, Sep 5, 2013 at 2:01 PM, Martin Thomson <martin.thomson@gmail.com>wrote:
> On 5 September 2013 13:43, Roberto Peon <grmocg@gmail.com> wrote:
> > That would be potentially problematic in some ways, as you really do
> want to
> > be in the http/2 context for the response.
>
> "That" being what specifically?
>
> The response is sent in HTTP/2.0, but the point was that the *request*
> is in HTTP/1.1.
>
> > We could solve some of this by stating that you must not have expect:
> > continue when upgrading to http/2.
>
> Roy's point here, I think, is that expect: 100-continue has to come
> before upgrade: foo, in all cases. We could prohibit the combination
> too, but perhaps there is no need. I suspect that anyone with any
> sense will try to avoid this combo anyway.
>