[JDEV] Napster and Jabber

[Eliot Landrum]

Evan -

Thanks for reviewing my proposal. Your comments are most welcome.

On Mon, 6 Dec 1999, "Evan Martin" wrote:

<snip>
 
> - Napster Server:
> "drscholl" (drscholl at hotmail.com) is currently developing a open source
> Napster server, and documenting the protocol as he does it.

Ah, that is interesting! Does drscholl's implementation provide backwards
compatibility or any special features?

> I try to keep the latest version of his protocol specification on my home
page
> (http://students.washington.edu/eeyem/), which is also where anybody
> interested can download my Napster client.  To just look at the protocol
> spec, visit http://students.washington.edu/eeyem/gn/napster.txt .

Yes, I browsed over it after jon at starbits.net mentioned it to me. From a
non-coder's perspective the only comment I can provide is that it still
wouldn't provide the flexibility that an XML based system might. Does anyone
else have comments as to why a Jabber implementation of this type of system
might be different?

> Napster uses multiple unlinked servers.  This means that the content on each
> server is independent of the others.  Napster servers are probably under
> incredible loads, because searching is expensive.

Ah! I wasn't sure if they were using one server or a distributed system. Do you
know if they use some type of standard SQL database? Are there plans to make a
distributed system that you know (at least in the open source implementation)?
I would like to figure out a way for Jabster to be distributed to decrease
server load. Suggestions welcome... ;)

> The Napster protocol currently has a few design flaws that prevent linking
> the servers.  This could be worked around, but it would break compatibility
> with the existing protocol.  The Windows client doesn't let you specify a
> server anyway, so that may be a nonissue.

Ok. One thing that I think a few people (I am not inferring that you have this
idea) that I have talked with about this have a misconception of is that
Jabster is not aiming to be a Napster-compatible solution. While we *COULD*
provide a Napster transport to translate between the systems, I was looking at
creating a completely new system that is, hopefully, superior.

> - Server Design:
> Eliot wrote, "- Server gathers list of results, checks to see which of the
> hosts that have
> the requested songs are online, pings them and sends list back to client."
> This would add the additional load of pinging to the server, which isn't a
> good idea and doesn't accurately represent the client-client ping anyway.
> I appreciate that the idea is new, I just thought I should add that you
> should offload as much processing as possible to the client.

Very true. It looked like this feature was in the Windows client and I thought
it would be good.. but after further discussion the Jabber system takes care of
that problem anyway with much less overhead.

> Finally, is there a mailing list archive where I could read the responses to
> his post?

Not many people have posted responses yet, but you can read the archive at:

http://mailman.jabber.org/pipermail/jdev/1999-December/date.html 

> (To the napdev list members: jabber is a messaging client, visit
> http://www.jabber.org for more information.)

It is a quite a bit more than a simple client, otherwise I wouldn't be
suggesting this idea. 

Thanks for the input! I'd love to hear more from you!

-- 
Eliot Landrum
eliot at landrum.cx