> From ivan.lovric@cnet.francetelecom.fr Wed Aug 25 03:46 EDT 1999
>
> >As for the distinction between caching and replication, I define
> >caching as creating a copy of an object at a point that would see
> >requests for the object regardless of whether or not it has a copy.
> >Examples: main memory cache on the browser, a proxy cache,
> >main memory cache at the server. That is, caching creates copies
> >along the normal path of request and data flow.
>
> Normal path for a request needs to be clarified :
> For instance transparent caching modifies the normal path of a request.
> Also, when using ICP with siblings, the request may be redirected to a
> sibling cache
> containing the requested URL, thus modifying the normal path.
>
Yes, clarification is in order, since one more person had essentially
the same comment.
Given the set-up, the normal (i.e., common-case) data path
does go through the cache in your transparent caching example.
It's just like with any other cache - it
you do not have it the data does not flow through it. So, we should
consider the given environment (in your example, that includes the
redirector and the node that receives the redirects). Given the
environment, we ask ourselves whether the redirects recipient is a
cache or a replica server, and it is a cache because it is on the
normal path of requests: requests arrive at it whther or not it
has the requested object.
As for siblings, they are on the normal path for *some* requests,
and they acquire their copies as the result of these requests (this
is another clarification!). Thus they are caches.
Misha
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:26 MST