Re: ICP & HEAD requests

On Sun, 6 Sep 1998, Chris Wedgwood wrote:

> On Sat, Sep 05, 1998 at 07:09:17PM -0600, Alex Rousskov wrote:
> > Looks like what you really need is Cache Digests, not
> > ICP-with-a-head.
>
> Why?

Because Cache-Digests will do exactly what you need with no code
to implement and with no latency/processing overhead for your sibling.
 
> I've read the cache-digest stuff, and I'm not all that worried about
> wads of ICP traffic (internally anyhow, externally I have other
> ideas), I use multiple NICs in the machines and they peer on a
> separate network segment where possible.
>
> So, why, in theory, can't we have `ICP-with-a-head' ?

In theory, we can have ICP-with-anything. :)
In practice, I do not see a good reason for implementing something that is
already supported with a better protocol. This is just my opinion, of course.

Alex.

P.S. Also note that including HTTP headers in a UDP transmitted message might
cause nasty bugs (besides extra disk I/Os required to fetch the headers).
Received on Tue Jul 29 2003 - 13:15:53 MDT