On Thu, Dec 04, 2003 at 05:18:14PM +0200, Costas Dokolas wrote:
> > -----Original Message-----
> > From: Niklas Bergh [mailto:[EMAIL PROTECTED]
> > Sent: Thursday, December 04, 2003 4:49 PM
> > To: 'Discussion of development issues'
> > Subject: RE: Re[2]: [freenet-dev] black hole report
> > 
> > > -----Original Message-----
> > > From: [EMAIL PROTECTED] 
> > > [mailto:[EMAIL PROTECTED] On Behalf Of Costas Dokolas
> > > Sent: den 4 december 2003 15:31
> > > To: Discussion of development issues
> > > Subject: RE: Re[2]: [freenet-dev] black hole report
> > > 
> > > > -----Original Message-----
> > > > From: Niklas Bergh [mailto:[EMAIL PROTECTED]
> > > > Sent: Thursday, December 04, 2003 4:03 PM
> > > > To: 'Discussion of development issues'
> > > > Subject: RE: Re[2]: [freenet-dev] black hole report
> > > > 
> > > > 
> > > > > - What could be the benefits of "routing-only" nodes? (i.e.
> > > > > low storage space, assuming they are on very fast 
> > > > > connections) Perhaps extra anonymity for the network?
> > > > 
> > > > A big problem here is that data still has to pass through 
> > > the routing 
> > > > nodes.. Routing nodes has to update their routing table, 
> > > right? And in 
> > > > order to update their routingtables correctly they have to 
> > > verify any 
> > > > data sent from their routing destination nodes. Same 
> > problem as in 
> > > > path folding, there is a need for some kind of alternative 
> > > > verification method.
> > > 
> > > I don't quite get it...
> > > 
> > > You mean that we need to keep transient data in order to 
> > > verify the CHK upon d/l completion? (and then some more until 
> > > the requesting node receives it, of course)
> > 
> > It is verified on the fly really but you have got it right.
> > 
> > > If you mean that, I was under the impression that we use temp 
> > > space for that anyway. On the other hand, I'm not dismissing 
> > > storage altogether, I only mean that it could be a of a size 
> > > enough for this kind of routing. In that respect, we don't 
> > > even need a store size limit, just a way to always delete 
> > > data after we're done forwarding it.
> > 
> > Every single data item that is cached by the (almost) 
> > routing-only node
> > will relieve one or more other nodes of sending that particular item.
> > Hence, the more cache on a node, the better for the network, 
> > always.. As
> > it is now.
> > 
> > What I meant was that the only way a non-storage routing node can be
> > really useful to the network if it allows the requestor and 
> > the located
> > datasource to transfer the data directly between themselves.. And that
> > process is as far as I understand called path-folding.
> > 
> > But the problem with this is that the routing node still has to verify
> > that the datasource really had the data it said it had in order to
> > reward that node in its rt. Because of that some kind of verification
> > mechanism which doesn't involve access to the actual data has to be
> > conjured up before routing-only nodes can be useful.
> > 
> > Clearer now?
> 
> Yeah, but...
> Why fold? I mean, isn't there real value for a node (many nodes if possible)
> that knows a lot about where to find data and who's best to get it from at
> the mere expense of a hop? On that area, isn't there value for any node
> (i.e. normal node) to know of more possibilities of where to find a piece of
> data?
> 
> Folding is only needed if the routing-only node is too slow (i.e.
> overloaded) to do the transfer itself, and I believe there are people (and
> non-people, i.e. institutions) willing to donate bandwidth much more than
> space, especially considering legal implications of actually storing info,
> whether they know anything about it or not!
> 
> I'm only getting into this line of questioning (besides personally liking
> Socrates' style of getting somewhere with Q&A; I don't know the English word
> for it) because routing, which is basically a "who has the data?" in
> combination with "how can I get it ASAP?", is the problem du jour.
> 
> Consider the following (maybe stupid):
> Until now, we assumed that a node passing info (or having just dropped a
> piece of info) need only keep a reference to the node it got it from. Does
> it still work that way? What if it also keeps a reference to the node it
> passed the info to? Isn't that a possible future source for the info? What
> if it keeps references to all nodes it got the info from and passed it to?
> That would work only for well connected nodes, and probably only if the info
> was close enough to its specialization to get more requests for it, but then
> again, it would hang on to the data, but consider instead the routing-only
> node...

It's a tradeoff between availability and specialization. The node that
made the request will probably be less specialized than the one that
served it - and we have no reason to trust them; it's a trust issue
really. Nodes are in our routing table because we trust them, because
they have served requests successfuly.
> 
> l8r,
> 
> Doc
> _______________________________________________
> Devl mailing list
> [EMAIL PROTECTED]
> http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.

Attachment: signature.asc
Description: Digital signature

_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to