Re: The End Of The End-To-End Argument

Date view Thread view Subject view Author view

From: by way of Rohit Khare (hallam@ai.mit.edu)
Date: Tue May 02 2000 - 19:54:11 PDT


I agree with much of David's argument, except that he leaves the erroneous
impression that the people demanding changes in the Internet architecture
are Web people. I am not aware that this is the case.

If Akami cannot manage their cache infrastructure without warping the
Internet it kinda brings into question a certain market cap n'est pas?
Perhaps they should go talk to Mr Rohit Khare whose scheme to speed up
caching was an order of magnitude more cluefull than their own.

I am reminded of the period in which ideas were tossed into the Web
infrastructure with zero thought such as Netscape's cookies, Sun's Java and
worst of all, the joint production, Javascript. Cookies were thrown in to
'solve' the 'problem' of online shopping without the slightest consideration
of privacy or security. The mitigating mechanisms were all afterthoughts. It
was a pity that the idea was published in code. Java is relatively harmless
insofar as practically no-one uses it for the touted 'active content'
application, but folk who complain about 'code bloat' should surely question
whether they would miss the couple of megabytes of disk it takes up and the
complexity it introduces. Finally I still haven't seen a javascript program
that did much more than solve one of the limitations to HTML identified in
1995 or earlier that in total would only require a few thousand lines of
code to fix. Instead we have an interpreter executing imperative code built
into every browser using a half-baked language the designer himself admits
to throwing together in a week.

Anyone building a network on the basis of complex routers rather than simple
packet switchers should bear in mind that they bear a considerable risk. If
they correctly anticipate the future evolution of the networks they win a
medium gain. If they do not anticipate that evolution they lose everything.
They will be stuck with a plant optimized for assumptions that are obsolete.
It is the RISC vs CISC argument. You can always squeeze a bit extra out of
the pot with CISC but each additional increment of complexity slows you down
when you attempt to adapt to the new environment.

         Phill

For help on using this list (especially unsubscribing), send a message to
"dcsb-request@reservoir.com" with one line of text: "help".


Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue May 02 2000 - 19:55:55 PDT