Re: [hybi] #1: HTTP Compliance

[Willy Tarreau <[e-mail]>]

On Thu, Jul 22, 2010 at 01:22:07AM +0100, Jamie Lokier wrote:
> Willy Tarreau wrote:
> > Hi Jamie,
> > 
> > On Thu, Jul 22, 2010 at 12:38:34AM +0100, Jamie Lokier wrote:
> > > Willy Tarreau wrote:
> > > > On Wed, Jul 21, 2010 at 08:22:11AM -0700, Roberto Peon wrote:
> > > > > I believe Mike quoted numbers in an earlier thread.
> > > > > 
> > > > > Just to be sure we're talking about the same thing, I'm talking about the
> > > > > UPGRADE feature of HTTP generically, and not as it applies to the current
> > > > > draft Websocket spec.
> > > > 
> > > > I thought they were related to the current version since it also uses
> > > > Upgrade. Anyway, even if some don't support it yet, it will be easy
> > > > to detect bypass them based on simple rules.
> > > 
> > > Um, will it?  How would that work?
> 
> [Good description of transparent proxies at ISPs with configurable
> HTTP-aware rules on the routers.]
> 
> Ah, I thought you meant "It will be easy to bypass *broken transparent
> proxies* based on simple rules *in the WebSocket clients and/or servers*".

OK.

> I.e. how to cope when the request hits a transparent proxy that
> doesn't work and you're *not* the ISP.

This happens all the time for such ISPs with simple HTTP sites that don't
display a video or something. They know about that very fast through their
forums from complaining customers, and they fix the issue using one of the
bypass rules above. And sometimes they ask the cache editor to fix their
product because they're fed up with feeding the L7 switch with bypass rules ;-)

There are not that many ISPs in each country, I mean there are far less ISPs
than there are web sites or potential WebSocket implementers. There's a high
pressure on them to work as expected by customers.

> Failing as fast as possible, and recognising it as a failure, is
> highly desirable if we can manage that.  I don't know if we can.


That's my precise intent : fail ASAP if we know for sure it cannot work,
and don't fail for no reason when we have no proof it cannot work when
it would most likely have. Please see my proposal in another mail about
that, I think it can improve on the current handshake for that.

> It would need actual experiments on a broad cross-section of the net.

I wholeheartly agree with that. It's the first time I read about testing
before specifying on this list ! Doing so from the beginning would have
saved a lot of man-hours !

Regards,
Willy