Re: HTTP/1.1 & Proxies
Roy T. Fielding (fielding@kiwi.ICS.UCI.EDU)
Wed, 02 Jul 1997 13:10:42 -0700
>>In general, there is no way to modify HTTP such that an invisible
>>broken proxy can be detected.
>
>Is this really true? It seems a very sweeping statement.
It is true provided that the proxy remains invisible.
Note, however, that HTTP/1.1 requires that a proxy be visible.
We have done all that we can do from an application protocol standpoint.
Note also that I distinguish between a proxy and a tunnel in the spec.
The problem with Gauntlet is that it is either a broken proxy
or a really lousy tunnel, and thus it is not interoperable with other
HTTP applications. In this case, Gauntlet is only visible when
a chunked encoding is used to delimit the message length, at which
time it screws it up. MSIE wants to detect it *before* it becomes
visible, which is difficult.
.....Roy