(file sharing) morpheus rules!

[Adam Back]

On Tue, Aug 07, 2001 at 09:27:42AM +0000, Steve Mynott wrote:
> Adam Back <adam at cypherspace.org> writes:
> 
> > If any were looking for a replacement for napster since
> > it buckled to pressure from RIAA, it's here: morpheus.
> 
> Are you sure this isn't just the first flush of enthusiasm which will
> fade in the cold light of morning?
> 
> Isn't it just yet another closed source P2P network like Direct
> Connection?

It is closed source and closed protocol.  The closed source issue is a minor
problem, someone will reverse engineer and re-write it, or design a new
incrementally better file sharing protocol.

The reason for enthousiasm is the demonstration of feasibility of a number
of open questions in scalability and performance which arise after observing
gnutella.  The fact that morpheus has reached the level of scalability it
has, speed of searches, and download speed is interesting because it gives a
lower bound on what's possible.  That it is possible is good news for the
future of file sharing.

> > Gnutella wasn't but didn't scale, morpheus appears
> > to be scaling and if anything performance is improving
> > as the data density gets higher so the sharing surface
> > can give you the content you want from closer and closer
> > nodes.  There may be an inflection point where it
> > starts to really take off as the number of users
> > is still improving the usability, performance and
> > variety of content.
> 
> Why don't you think Gnutella doesn't scale?

Empirical evidence gathered by trying to use it over a number of months.  It
starts slower, searches are slower, and downloads fail a lot (~90%) of the
time, resumes don't work as well, resumes can't switch source, it doesn't
download from multiple sites simultaneously, plus morphus has a number of
really good client GUI features which positively affect the richness of
meta-information available for searches.

> <http://www.openp2p.com/lpt/a//p2p/2001/01/25/truelove0101.html>

While some of the excuses given here are undoubtedly valid, the primary
problem -- broadcast searches -- is not addressed except by a not widely
used and manually configured fast host acting as a proxy shielding a slow
host from search requests (if I interpret that right).

Morpheus on the otherhand automates it's solution to the general issue of
different host speeds and scalability of searches -- self organising
selection of super-nodes -- and does many many other things better than
gnutella, and discards it's speculated intra-version compatibility issue
effect on performance by virtue of there currently being only one version
(or more properly the versions all rely on the same comms layer library).

Adam