[JDEV] (no subject)

Pat Magnan pat at sluggo.org
Mon Jan 27 19:21:45 CST 2003


At 04:13 PM 1/27/2003 -0700, you wrote:
>Daniel MD wrote:
>
>>hello, i was wondering what is wrong with the protocol that it can't take 
>>dynamic IP's ?   Jabber server, thus moving Jabber closer to a 
>>peer-to-peer model (currently this would require each device to have its 
>>own fully qualified domain name). I really would like to implement a p2p 
>>jabber network.

So does the Internet. If you want to change that, take it up with the guys 
who built it ;).

You can probably setup the jabberd server to be called 'localhost' and make 
the clients connect by IP address, why would you want to.. Every client is 
a server, then who gets to be tom or john? How do you route messages with 
such chaos?

Napster et all don't work completely dynamically either, someone somewhere 
controls who you are (server), and says you are johny28382, and then 
figures out you want to talk to susie12 and connects the two IPs directly. 
That server had to be registered in DNS (or have a fixed IP).

Jabber isn't crippled this way. No one cares who the client's IP is, only 
the server must be a known entity. I use Jabber for what is essentially P2P 
communications right now (two instances of my program exchange information 
over jabber), without issues.

If you want p2p file sharing, there's other things that accomplish that 
that are GPL and you can build what you like on top of them. Or use Jabber 
as the server, and the 'clients' can exchange 'file list' type messages 
with multiple other clients simultaneously as a result of a search. 
Personally, that's over engineering an established solution in my books, 
I'd grab a GPL'd p2p program, and extend it rather than kludging an IM 
solution to do it. But, that's my preference.

Jabber does exactly what its been designed to do. And is flexible enough 
(with sufficient work on the developer's part) to do more. Losing the 
'authoritative' server won't make it more flexible imho :).




More information about the JDev mailing list