Re: Microsoft CARP

On Sun, 12 Oct 1997, James R Grinter wrote:

> On Sun 12 Oct, 1997, Garry Glendown <garry@insider.regio.net> wrote:
> >Paul Chinen wrote:
> >> What I think is that MS does not understand quite well ICP:
>
> [the conspiracy theory says that it's because they can't implement ICP in their
> server model. A colleague pointed out that it also looks impossible to
> implement the CARP hashing in javascript.]

I'm a little bit confused with all this CARP ;-) It's supposed to
eliminate ICP query roundtrips which seems fine to me in the case of huge
delays. Huge delays are common in wide-area configurations where it is
reasonable for network managers to require some (or full) control over
what&where data is cached (think of *.edu or the Japan Cache project).

Back to local area configurations, using cache arrays seems to be a good
idea but I don't get where the benefit is comparing to a DNS-based
solution and whether the difference in performance is noticable (given
that bandwidth is plenty - ICP comes for free).

I also don't see why we need yet-another-protocol as CARP functionality
can be incorporated into ICP (hey, there's plenty of room there)....

As for eliminating ICP query delay, one could distribute some sort
of "indexes" in a space-efficient form which could be referenced to
resolve next-cache-hop. A probabilistic approach (with a high confidence
factor, of course) could be the answer to this. Has anybody looked into
this issue?

... or maybe I'm totally confused ;)

-K.
Received on Sun Oct 12 1997 - 12:12:06 MDT