Lists Home |
Date Index |
On Wed, 23 Jan 2002, Paul Prescod wrote:
> My point was, it isn't right to say that "UDP is reliable", "HTTP is not
> reliable." Nothing is completely reliable (because the physical universe
> isn't completely reliable). Anything can be made "as reliable as
> possible" (given the above limitation). The only question is how much
> extra work it takes. UDP apps have to do a lot more work than TCP apps,
> which in turn do more work than HTTP.
Yep, I'd agree with all that. However, the problem is that TCP does *too
much* work for many applications :-)
> We aren't implementing DNS. Doesn't it seem a little odd to post about
> the inefficiency of TCP to an *XML* mailing list? Where is the
> inefficiency of a SOAP call really going to come from?
All sorts of places! But somebody somewhere said that HTTP was great since
it had all these features, and I felt compelled to point out that many of
these features are tradeoffs taken to implement serial lines, is all...
PER over UDP will be more efficient in bandwidth usage than XML over UDP
which will be more efficient in latency than XML over HTTP, special cases
And it's a shame that people are making decisions that currently mean that
Web services won't be able to be as responsive as DNS no matter how well
the server is written...
> Paul Prescod
Alaric B. Snell
http://www.alaric-snell.com/ http://RFC.net/ http://www.warhead.org.uk/
Any sufficiently advanced technology can be emulated in software