Re: pipelined client/server behaviour

On Apr 1, 2007, at 11:44 AM, Jamie Lokier wrote:
> Roy T. Fielding wrote:
>> But there is no question that
>> pipelining is hard to do right across all implementations.  That
>> doesn't mean it can't be done across some implementations, and there
>> really isn't much point in discussing it -- the standard will not
>> be changed.
>
> Without changing the standard -- here's a proposed enhancement:

That would obviously change the standard -- in this case, by implying
that all the existing implementations of pipelining, the vast majority
of which do work quite fine, do not work just because they lack this
new header field.

There is nothing you can add to the protocol to fix non-compliant
intermediaries and servers that are already deployed -- no matter
what you do, they will behave in some non-compliant way to screw up
the plan.  The right thing to do is let breakage occur until someone
gets off their rear end and replaces the broken software with
something not so broken, at which point the existing protocol
defined by the existing standard, which has already been proven
to be interoperable, will work fine just as it does now with the
majority of servers already deployed on the Internet.

....Roy

Received on Monday, 2 April 2007 12:53:07 UTC