Re: Minutes / Notes

[Keith Moore <moore@cs.utk.edu>]

many p2p apps do their own routing because they are forced to do so, 
not because this is a desirable or acceptable burden to impose on
applications.    also, what works acceptably well in a system to
transfer widely-replicated files does not necessarily work well in a
system where host identity is important.

Keith

> It's worth noting though that this is what these p2p applications do, 
> it's the core of their development. one could equally argue that 
> anykind of ad-hoc mesh protocol work could learn a lot from the p2p 
> apps. it's a long way from that to expecting average program to do the
> same things.
> 
> > I heard many say during the multi6 debate that applications cannot 
> > possibly handle multiple addresses well. To disprove that, I suggest
> > you take a look at the variations of "swarmcast" implemented by P2P 
> > systems such as Kazaa or E-Donkey. These systems are about
> > exchanging files, so they have their own identification system --
> > file names. The naming system can resolve file names to multiple
> > locations of a file. The transfer system then asks slices of the
> > file from a set of more or less randomly chosen locations.  It
> > automatically adjust to topology by asking more from those locations
> > that serve faster. And it automatically compensate connection losses
> > by just repeating the slice request, or a fraction of it, to some
> > other nodes. Such a system would handle multi-addressing
> > beautifully.