Re: Microsoft CARP

On Mon, 21 Jul 1997, Redfern Ian wrote:

>
> They have moved on from the discussion in
> <http://ircache.nlanr.net/Cache/ICP/draft-vinod-icp-traffic-dist-00.txt>
> to an implementation in their next release.
>
> Should Squid support CARP?
>
> Ian Redfern (redferni@logica.com).
>

 I knot it's stupid to talk about something before reading it but... Whats
the big deal about a set of peers having duplicated data? So it wastes
some drive space and memory.. The only situation where I can see doing
that as a win is where peers are tightly coupled (network wise, like the
nlanr caches) and there is little penilty for fetching from another one..
And even there simple rule blocking will suffice... This method just means
that if a single servers goes down then there will be no caching of some
specific URLs.. Further more, if a browser must connect to multiple caches
within a single session, that gives it less of a chance to use http/1.1
keepalives and such...

 So, to me it seems that CARP isn't that useful and it's own overhead may
make it useless in most circumstances.. Protocols like that are only
useful in commertial products without source. If your needs were ever so
strange that you 'had' to have that feature you could easly pay someone to
add it to squid (and then of course give it back to the rest of us as a
compile time option)..
Received on Mon Jul 21 1997 - 05:28:13 MDT