[Tsvwg] Re: [e2e] e2e principle..where??....

[David P. Reed]

At 11:57 AM 6/5/01 -0700, Eric A. Hall wrote:

> > Thus, a middle-box-cache that tries to spoof both ends will almost
> > certainly get the answer wrong sometimes.
>
>...which is fine as long as the participants are willing to trade veracity
>for performance, *AND* they have the opportunity to bypass the middlebox
>when they need to sacrifice performance for veracity.

With a true end-to-end design, you can get veracity *and* performance.

The problem with veracity in the kludge-caches is that they aren't part of 
the application layer, so they can't do the right application-dependent stuff.

By generating pages with URLs that point to caches closer to the browser, 
you can "distribute" web service quite effectively and gain the faster 
fetch time, and even distribute the processing cost over a lot of 
caches.  This is something that can't be easily done by mere "protocol 
spoofing".

The idea that "performance" always suffers when you follow the end-to-end 
principle is just plain false.






- David
--------------------------------------------
WWW Page: http://www.reed.com/dpr.html