On Tue, 26 Dec 2000, Timm Murray wrote:
> While date rewriting is a pretty good kludge (as kludges go), I think it
> will be harmful to routing over the long term. Take this simple example:
>
> A-B-C-D-E-F
>
> E inserts such a document with htl 2 and sends it to D and D sends it to C.
> A is looking for that document, sends it to B, B sends it to C, finds the
> document and sends it back through the chain. Now both B and A have it.
> Tommorow, E inserts the updated document, again to D and C. Now the copy on
> A and B is obsolete (and will eventualy drop out of its store) and A will
> have to get the new document by requesting again. Thus Freenet's "routing
> gets better over time" claim is undermined, as the routing only has one day
> to get better.
The updated document would probably not go to node c, because node c is
probably not the closest node to the new document. If you're basing your
premise on that false assumption, you're wrong. Changing the date in the
SSK changes the routing key.
Actually, frequently updated keys will help Freenet organize itself, I
think. You're guaranteed a lot of new documents to work with. New
documents are good.
Ultimately, though, I don't think it matters much at all.
> This example shows that while date-key based updates are fine for now, we
> need a real updateing solution to replace it.
Like what?
> I beleive I have an error in the example above in that A and B don't know
> about any referances to the document and thus can't route the request (I'm
> not too sure on what Freenet does with a request when it doesn't have
> routing information). In any case, I beleive my point has gotten across.
A and B have references to node C or D, which specialize in data with keys
similar to the document requested.
--
Mark Roberts
[EMAIL PROTECTED]
_______________________________________________
Freenet-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
