Re: Hash algorithm in javascript

Sorry, but CARP is very much a client thing.

CARP is a hashing algorithm for request routing combined with and a
standard exchange format for exchanging weights to be used in this
hashing algorithm when contacting a cluster of caches. In the intended
scope for CARP the hashing/routing algorithm runs in the clients, who
download the weights table from any of the array members.

What is missing from the "standard" to gain acceptance is

a) An array membership registration method. How the array members
registers into the array is completely outside the specification, making
it impossible to build a CARP array of caches from different vendors.

b) Direct CARP routing in the browsers. Is today implemented via PAC
scripts, but PAC browser support is still somewhat flaky..

c) A good method of distributing the location of the PAC file to the
browsers (i.e. WPAD or a similar thing).

Because not all clients can be expected to support or know about CARP,
CARP also includes intra-array routing where array members re-route
requests arriwing at the wrong array member.

Don't confuse CARP request routing, and CARP intra-array routing. The
first is CARP, the second is a bandaid to cope with braindead clients or
temporary hickups.

No, Squid does not implement CARP, only the hashing algorithm of CARP.
Saying that what Squid has is CARP is false marketing as less than 1/4
of CARP is actually implemented.. (no intra-array routing, no
downloading and processing of CARP membership tables, no publishing of
CARP membership tables, and no CARP membership manager)

/Henrik

Chemolli Francesco (USI) wrote:
>
> > Use CARP. No need to write new scripts...
>
> CARP is server-side, I wanted to do something client-side.
>
> Basically I'm trying to replace CARP by .pac + cache digests.
>
> --
> /kinkie
Received on Wed Mar 21 2001 - 10:56:40 MST