Tue, 16 May 2000 13:27:19 -0400
> At 08:12 AM 5/15/00 , Tyler Close wrote:
> >Markm wrote:
> > > Why should I expect that the set of SLSs that my user agent
> > > queries will
> > > intersect with the set of SLSs that your site is
> registered with?
> >Because all SLSs in the web get the opportunity to store the
> >registration. In this way, SLS is more like DNS than VLS.
> Perhaps I'm so hopelessly confused I don't know what I'm
> confused about, so
> I'll ask a possibly stupid question. Are you assuming all
> site location
> registrations get replicated to all SLSs?
I was assuming replicated entries and trying to not think about the
large storage burden that this entails. I guess you're going to make
me think about it now.
Since the SLS is a web, and not a tree like DNS, fetching from your
neighbour/parent, as DNS does, is a bad idea. I guess this is why E
came up with location hints.
So with VLS-like location hints:
where ;<SLS-IP-address> may be repeated 0 or more times.
So the user-agent should send (location-hints, public-key-hash,
scheme) to the local SLS, which will return a list of (DNS hostname,
IP address, port) sorted by increasing ping time.
Using location hints instead of "broadcast and store" introduces a new
problem. When does the local SLS re-query the hinted SLS? This needs
to be more often than "none of the old entries respond anymore" if we
want to be able to introduce new mirror sites on-the-fly.
So far, the best I can think of is for the local SLS to re-query the
hinted SLS once the latency and/or throughput, of the best host falls
below some threshold, in effect, checking for a mirror when it wishes
there was a mirror. If there was a way for a site owner to be notified
when a particular SLS was pestering for a mirror, then this could be a
hook on which systems like freenet could be built.
Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger.