Re: [HTTPSubstrate-16] Propoposed criticsm of RFC3205

HTTP was designed for a class of applications.
Other applications which have very similar requirements
for latency, transaction semantics, security, error recovery,
communication of options, client and server capabilities, etc.,
might well fit as "reasonable" to layer on top of HTTP.

Unfortunately, most of the applications that layer on top
of HTTP are, unfortunately, very much unlike the application
that HTTP was designed for, and have very different requirements
for security, transaction semantics, error recovery, etc. etc.

There is a separate issue (which I've raised repeatedly), that
the common infrastructure for the web in use today includes
so-called "transparent proxies" on port 80, and these are
likely to interfere with use of the port 80 for non-browsing
applications. It would be inappropriate to attempt to design
a robust network protocol using HTTP over port 80 without taking
into account the likely operational environments into which
that protocol might be deployed.

Admittedly, there are many environments where such proxies don't
exist, but the fact that you might be so fortunate doesn't
relieve you of the obligation of accounting for them.

I don't think it's appropriate to make blanket comments
on RFC 3205 without addressing these issues.

Received on Friday, 7 February 2003 21:38:09 UTC