> From owner-wrec@cs.utk.edu Fri Aug 20 15:42:53 1999
> Delivery-Date: Fri, 20 Aug 1999 15:42:54 -0700
> Return-Path: owner-wrec@cs.utk.edu
> To: wrec@cs.utk.edu
> Subject: Joe Touch: Re: Taxonomy draft - web replication problems
> Date: Fri, 20 Aug 1999 18:42:25 -0400
> From: Micah Beck <mbeck@cs.utk.edu>
> List-Unsubscribe: <mailto:wrec-request@cs.utk.edu?Subject=unsubscribe>
>
> Joe Touch wrote:
>
> > > > The difference between a mirror and a replica, IMO, is that
> > > >
> > > > a mirror is a copy of an entire server
> > > > a replicate is a copy of one of the files on a server
> > >
> > > but then we run into problem of defining what is an 'entire server' ?
> >
> > "entire database" - a related collection of data.
> >
> > I.e., mirror refers to the server,or collection, not the
> > individual items. A datum is replicated; the collection is
> > mirrored. (at least that's what I'm trying to convey).
>
> I think the distinction here is "on-demand" vs. "a-priori" not whether
> the whole database (or channel, as we would say in I2-DSI) is copied.
> One can have a dynamic cache which fetches the whole channel when any object
> is referenced. Similarly a-priori strategies for copying can be based on
> factors other than "the whole database". The difference is whether it is
> a the copying is a response to a request or made prior to the request. >
Prefetching or push caches can grab the copies of an
entire server (assuming it's completely 'connected' via internal
links) a-priori. Isn't that all prior to the request?
I see the difference as a combination of all three factors
- ahead of time
- entire database
- permanent copy (source invalidated, vs. timeout in a cache)
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:26 MST