Licensing and the client/server divide [was Re: License formixminion dist]

[Nick Mathewson <nickm@alum.mit.edu>]

On Tue, 2002-04-23 at 14:57, Zooko wrote:
> 
> > > I think it would be best if it were under LGPL, ...
> [...]
> > Hn.  I'd just as soon _not_ help people make proprietary servers that
> > subsume the Mixminion server.  If they extend the Mixminion server, I'd
> > rather they released their code.  (Are there any realistic examples of
> > why we should permit this?)
> 
> Well, any system that uses MM for anonymous messaging would require linking 
> with the MM code, right?  Like, I dunno, some distributed database for some 
> kind of medical support group or something.

I don't think so.  See below....
 
> I still don't understand the distinction between "client" and "server" here.  
> To my mind every MM node is identical, and you must run a local MM node to use 
> the MM protocol directly.  (All this SMTP stuff is, in my view, just one 
> possible service that a person A can run, where person A is required to have 
> an MM node and then offers service via SMTP to a person B who doesn't run an 
> MM node.)

Ah.  We may have a mismatch here.  To my view, running a mini-node for
local delivery is a fine thing, but it should not be necessary to do
so.  In my parlance, a "Client" is a piece of code that talks to a
server, or prepares a message for a server, or processes a message from
a server, but is not a server itself. 

Client code will include: code to package up a message and send it to a
MM node that accepts messages from non-nodes; code to create reply
blocks; code to read messages sent with reply blocks; code to manage
Nymservers, and so on.

IMO, it should be possible have a system that uses MixMinion for a
transport layer without extending the server, in the same way that pine
doesn't need to extend sendmail.  Such a system can tell its (possibly
local) MM node that it has a message to send, or check a mailbox-like
location for received messages.  It shouldn't need to extend or modify
the MM server; instead, it can just use the client code to generate and
queue outgoing messages and process replies...

...unless it needs to add a new delivery mechanism, or change the
operation of the server code.  I'd just as soon that such changes were
contributed back to the Mixminion project.  That's where I'm advocating
the GPL, possibly with something like the AGPL's part 2(d).  

So, to rephrase my questions, do we ever want to allow people to add new
delivery mechanisms or change the operation of the server code, without
releasing their source--keeping in mind that proprietary programs will
still be allowed to insert messages into the system and receive replies.

Yours,
-- 
Nick