On Sun, Jun 03, 2001 at 10:05:36PM +0200, Oskar Sandberg wrote:
> Another reason for seperation is that you could make the resulting URIs
> protocol independant, to work for people who want to share files directly
> (ftp or http) as well as via Freenet. (For example one could already hack
> a gnutella servlet that works from a keyindex (I think).)
Hmmm, Gnutella doesn't have URIs AFAIK?
> I think that the difference between searching for data reference based on
> an idea, and finding actual data based on a reference, are sufficiently
> different that they should be in seperate layers.
Certainly separate layers, I have no problem with that - but I see
little point in making them separate applications.
> Either way, I have been
> thinking for a while that we should at least experiment about this by
> making it a seperate application for Napster/Gnutella type filesharing.
Can you define what you mean by this?
> Trying to simulate searching is pointless, so the best way would be to
> experiment on a platform that could be useful but without polluting the
> more "serious" codebase.
I see no reason to think that simulations are pointless, but I do agree
that we need to implement. It is about time I did some coding again, so
I plan to get started on this over the next few days. I have an
architecture in-mind, which will be suitable for whatever we decide to
do with it.
Ian.
PGP signature
