Re: p1: Receiving a higher minor HTTP version number

[Willy Tarreau <[e-mail]>]

On Sat, Apr 20, 2013 at 04:30:17PM +1000, Mark Nottingham wrote:
> 
> On 20/04/2013, at 4:29 PM, Willy Tarreau <[e-mail]> wrote:
> 
> > On Sat, Apr 20, 2013 at 02:07:17PM +1000, Mark Nottingham wrote:
> >> I don't see anything in p1 2.6 Protocol Versioning that explicitly says an
> >> implementation ought to accept a message that has the same major version
> >> number it implements, but a higher minor version number.
> >> 
> >> I think we need to spell this out, because IME some servers do error out on
> >> (for example) a HTTP/1.2 request.
> > 
> > Makes sense but I'm not sure that these implementations will change for
> > this these days anyway, with 2.0 coming. Also we have seen with the
> > 1.0->1.1 transition that the minor change was not that seemless (specifically
> > due to persistent conns).
> 
> 
> Yeah, if this is uncontroversial, I can see adding a sentence or two (maybe
> with a requirement); if not, it's probably not worth the time.

Maybe at least what we should (IMO) say is that if an implementation receives
a message in a minor version it does not support, it should apply the
semantics of its own version and respond with its own (and probably refrain
from using persistent connections). This would give guidelines to people who
just implement 1.0 in simple scripts (though it's unlikely that these people
read RFCs anyway).

Willy