[Discuss] A Web Server is No Longer a Web Server

[Dale R. Worley]

> From: Kent Borg <kentborg at borg.org>
> 
> A system designed for hypertext has turned out to be the architecture 
> for so much of modern life, but it wasn't designed for that.

If you mean by "system", HTTP, it was always a client-server
request-response protocol, and it works rather well for that.  Witness
that so very many things use it.

The secret of its success is a bit subtle, I think.  There have been
lots of protocols that do similar things, and many other protocols as
well.  But many of them are brittle and high-overhead to use.  As Scott
Adams once noted, there are millions of ways of configuring an ISDN
connection, and if the two ends aren't configured agreeably, the
connection doesn't work.

Out of the early Internet came the realization that you can't afford to
have the people at each end configure things, the two systems need to
come to agreement between themselves.  So over the years, the IETF
learned how to design protocols that have immense "upward
compatibility" -- a request may come with many bells and whistles, but
the server can safely ignore the embroidery it doesn't understand and
use the parts it does understand.

So as all sorts of new features were larded into HTTP, if you put a
little care into it, you can use those features with your correspondents
who know them and still do business with those who don't.

Dale