[Previous entry: "Oridean & Akopia"] [Main Index] [Next entry: "Groove review"]
02/21/2001 Archived Entry: "P2P ramble notes"
I had a ten-minute ramble with Campbell and Henrik about P2P. Here are a few random notes I've jotted down - feel free to come up and talk to me about any of them if you want more info.
P2P is just the latest version of what the Internet was built on
SMTP
IRC
NNTP
... but that was when all Internet-connected machines were full hosts with static IP addresses.
Now, there's a (mostly) clear division between clients and servers.
P2P is an attempt to take the users' power back from the servers.
The second stage for P2P on the Internet was in the mid-90's when instant messaging started up.
Systems like Hotline attempted to slightly redress the balance - a server-based easy way of getting chat and file-serving going on a machine. But it never took off much beyond its original Mac base.
ICQ, while depending on servers, uses peer-to-peer for message and file transfer. I find it a very useful way of transferring files most of the time, except when firewalls get in the way.
There were tons of short-lived ones... Activerse's "Ding!" was quite good and had lots of good ideas, but not enough to kill the ICQ snowball. ICQ also did a groupware client & server thing that isn't on their site any more, but James Cronin still has a copy.
Instant messaging is still a hot topic, as well as ICQ and AIM (now both owned by AOL) there are various Microsoft protocols and Jabber, an open source platform.
Microsoft's done various different IM systems and protocols:
MSN Messenger
IMPP and RVP - which is which? Are either used?
and now Hailstorm, which looks like it could be fascinating - using an IM platform as a more generic asynchronous message architecture for other kinds of data. (I want to do more looking in that direction, I think there's a ton of potential)
Jabber is something I have mixed feelings about - on the one hand it's doing very well for itself and is not going away any time soon, on the other I think it was built on fundamentally shaky foundations that may hold it back from being really good.
As for the current P2P stuff...
"P2P" is a really nasty buzzword. It does NOT mean peer-to-peer, not in the way it's currently being used, anyway. The two most commonly given examples of P2P apps are Napster (only partially peer-to-peer) and SETI@home (not peer-to-peer at all - it's pure client-server). The whole thing reminds me of the "Push" fiasco of '97.
What P2P actually means is a method of locating machines on the client-side of the internet. The method we usually use to locate internet hosts is DNS, but that won't work for most clients. So P2P schemes set up their own naming and location methods.
For more on this, see my chum Clay's article about this. Clay has written a lot about P2P. He gave the keynote at O'Reilly's recent P2P conference. O'Reilly run a great site about P2P issues called OpenP2P. Check it out.
So, what P2P is meant to do is solve the location and transaction problems involved in client-to-client transfers. Location is only half the problem - even when you know where someone is, there could be a firewall in the way. Different programs use different methods of getting past firewalls with varying degrees of success.
Napster is, in my opinion, a great idea with a terrible implementation. In this case, the idea has got past the fact that the software is horrible. However, Napster the company is doomed. They will not make money unless they radically change their proposed business model. Users with any sense will switch to a free service such as OpenNap or Gnutella - after all, the whole appeal of Napster is getting something for nothing. (See the comment I posted to this Slashdot article - it's the first one)
There's better software than Napster for music swapping. Of course, to beat Napster, it can't just be better - it has to be twice as good, because that's the only way to overthrow something of Napster's popularity. My favourite is Audiogalaxy's Satellite for many reasons - I'll give you a demo if you like. I completely love it.
There's also Gnutella, of course. It's a lot friendlier than it used to be now that clients such as Bearshare have appeared, though I think it'd still be a little daunting for most. The trouble is that Gnutella's current protocol, while being a nice idea, doesn't scale well. There is work underway to fix this however. And there are loads more clients coming. There are also other Gnutella-like systems such as Kazaa working on the problem.
Then, of course, there's the whole P2P groupware thing. This is where Groove is. This kind of thing has actually been around for years - it's just a newer version of Hotline. There's also Thinkstream's "Tadaaa!" which is one of the worst product names I've ever heard.
Some other triggers:
* Metcalfe's Law
* Infrasearch
* Give & Take as file commands
* The Cornucopia of the Commons
Come up to me and I'll tell you about them!