--- Tom Kaitchuck <[EMAIL PROTECTED]> schrieb: 
....
> 
> If your talking about one idle node sending
> advertisements to another idle 
> node, why not just exchange data? Then the data
> could be moved to a place 
> where it is more likely to be requested.

Not a bad idea either.  It would be a way to populate
newbie datastores.  Still my idea was just exchange
the ads since, that could be done much quicker.  I
can't move 8GB of datastore, but I could move 16MB of
ads.  You get about the same effect, just one more
hop.

When you talk about swapping actual data around it
kind of reminds me of free haven.  They worried about
people swapping for stuff and killing it.  I'm not
sure if that's any more a problem here though than
some node dropping requests to censor.

> 
> It could work like this: If a node has been idle for
> a while, it looks in it's 
> store, picks the key furthest from it's
> specialization and sends it to the 
> node that it would request that data from if it were
> to do so. (Unless it 
> thinks that node is busy) Then that node, can ether
> reject the data, or store 
> it, and return it's least recently accessed data in
> return.

Hmm, seems like you'd have a slow motion reinsert of
all data.  The objects should eventually meander to
where there wanted.  So can I trust the last accessed
date of the object I'm given?

Oooh, your idea there gives me a big one.  Could the
probablistic caching be modified so that a node will
more likely store data within its specialization. 
Data with your specialization if the "evil ones"
examined your store would be harmless.  Then again it
might give them a better idea of stuff cached because
you requested it.

We have to balance caching things recently requested
with stuff in my specialization.  We could drop stuff
that's out of our specialization before we drop stuff
in it.

__________________________________________________________________

Gesendet von Yahoo! Mail - http://mail.yahoo.de
Logos und Klingelt�ne f�rs Handy bei http://sms.yahoo.de
_______________________________________________
devl mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to