Brainstorm: p2p client

I remeber that Naptella thread from long ago and it stuck out in my mind because that is when I realized that p2p networks must have a social conscience in order to succeed. There are a few techniues that must be employed that connect people to people along with their libraries. While TankGirl was wrankling with the behind-the-scenes protocalls that would allow a serverless network to exist, I wondered about the superficial features that users would see when they used the software.

At the time TankGirl presented her idea Gnutella was both fairly new and fairly usable. Since then that network has deteriorated (especially for us 56Kers). The funny thing about Gnutella's interface is that it's guts are plainly visible in the UI. Most of it's code is dedicated to maintaining the network rather than allowing convenient filesharing. So TankGirl suggested a p2p network that would work kinda like Napster and kinda like Gnutella, serverless yet centered around filesharing.

Well I think the main part of maintaining a community is communication, and that is what Napster really had going for it. Serverless networks don't facilitate that kind of communication because their users can't 'meet' on a central server that's always connected. In this respect I think that the old Hotline system was better than serverless p2p. With Hotline, like FTP, you would connect to a small server that hosted about 20 or 30 users who uploaded and downloaded mp3's to and from that server. The administrator would decide what files would be shared and discarded and organize the files into thematic directories. The things that distinguished it from FTP were the news, chat, and IM functions. The admin would periodically post messages to a news file, and any user could talk to any other user on the server. I was attracted to the servers by the music libraries that I could browse, every once in a while I would find a song I had forgotten about. I liked chatting too. Now I think that it's time to resurect the Hotline idea and make it serverless.

To that end there needs to be a few features in a p2p client's interface that not only facilitates but also promotes communication between users. A p2p program that allows library browsing should have a chat screen integrated into the hotlist window. When you hotlist a user, you connect to that user's personal chat channel and can talk to that user and anyone else who is connected to that channel. Naturally people will hotlist users with the largest libraries (who also happen to have the most bandwidth) and people will hotlist libraries that have music they like. As a result people who like the same music will easily find each other.

The hotlist/chat window will obviously contain a list of users that are currently connected. The program should list the songs that the users are currently listening to beside their names in that list. This would accent the the way people browse. When you see that someone is listening to a song you've never heard you can ask them what it's like and how much they like it.

When a couple dozen users are connected to one user they can search and hotlist each other's libraries. The network would be small and the total number of files low, but everyone on the network would share the same musical interests so finding good music would not be difficult. This 'pocket network' would be the foundation of relationships that would last and a community that stays together. Eventually the pocket networks would merge as they find each other.

I think this constitutes a protocall that would support a serverless network on its own. Each node would have the ability to act as file hosts, chat servers, and search engines, fulfilling the basic needs of a p2p network. When two nodes connect together they trade hotlist information. This allows the two users to search and browse each other's files. Then when more nodes connect they exchange hotlist info with the node with the most connections, or the mother node. The mother node would mintain an image of the network that it would distribute to all other nodes, that way if the mother node disconnects all the other nodes would have the information they need to reconstruct the pocket network.

The pocket networks would necessarily be small to avoid the problems Gnutella is having, but a small network wouldn't be too bad for the reasons I described above. The main obstacle would be finding other users to connect to, but Aimster has solved that problem. AIM and ICQ can provide the IP inforamtion needed to start a new network or find an existing network, and users would add the people they find to their buddy lists. The client would import the buddy lists from AIM and ICQ and make them available to others on the network. As users move around from one pocket network to an other the overall network would constantly change and there would be no shortage of new and interesting songs to share.

[TankGirl]

Hi Mazer, a good post with many interesting points. So let's dig into it...

Mazer:
I remeber that Naptella thread from long ago and it stuck out in my mind because that is when I realized that p2p networks must have a social conscience in order to succeed.

Social conscience (or consciousness) is not necessarily a precondition for the success of a p2p network. There is a place for 'asocial' networks as the relative popularity of Gnutella and Morpheus shows. Many people are happy just to leech the stuff they know to exist from anonymous sources and leave the entire social aspect out of it. But those who have tasted the possibilities of a social p2p environment know better and given a choice will gravitate towards networks that enable the exciting prospects of human-to-human communication.

Mazer:
Well I think the main part of maintaining a community is communication, and that is what Napster really had going for it. Serverless networks don't facilitate that kind of communication because their users can't 'meet' on a central server that's always connected.

This is true about existing serverless networks but in no way an inherent limitation of the serverless p2p architecture itself. 1-to-1 communication already works on WPNP and there are no principal obstacles to developing software that would allow arbitrary peers to establish 'meeting points' for chatroom-type activity either on their own or in co-operation with other peers. It is also possible to establish a reliable mechanism for two arbitrary peers to find each other on a pure p2p network. This is an area where the infrastructure of Morpheus shines in comparison to Gnutella and WPNP. With the additional 'metalayer' of connected supernodes Morpheus is able to bind the entire community into one reachable whole whereas Gnutella and WPNP can offer only a limited visibility and reach for each peer.

Mazer:
In this respect I think that the old Hotline system was better than serverless p2p. With Hotline, like FTP, you would connect to a small server that hosted about 20 or 30 users who uploaded and downloaded mp3's to and from that server. The administrator would decide what files would be shared and discarded and organize the files into thematic directories. The things that distinguished it from FTP were the news, chat, and IM functions. The admin would periodically post messages to a news file, and any user could talk to any other user on the server. I was attracted to the servers by the music libraries that I could browse, every once in a while I would find a song I had forgotten about. I liked chatting too. Now I think that it's time to resurect the Hotline idea and make it serverless.

Your description of Hotline was very interesting as I have never used Hotline myself. Small, dedicated 'clubs' of trusted peers sharing exclusively among themselves is the natural next level above the grass-root level of 1-to-1 communication in the social self-organization scheme of p2p. Here is my related comment to Jack in a discussion that took place recently on ZeroPaid's WinMX Forum:

=======

JackSpratts:
so you've got 1/2 gig per user. not bad. i wish morpheus had that much.
remember the old days? the only limit was your hd. i used to keep 20 gigs on 18 hrs a day, and so did a lot of people i knew. even more sometimes. but that was back in opennap time. i wonder if p2p will be able to handle those kinds of numbers in the future because now when i put that much up on win/bear/morph it's goodnight irene.

My view is that p2p will be able to handle even very large libraries but not in the flat, open-for-all fashion applied by current clients. The index information for a large library is a respectable chunk of data in itself. You want to minimize all unnecessarily moving around of such chunks so clever caching schemes (supernodes etc.) are a must. Even then, each time a large library is browsed - and browse we want! :-) - such a chunk has to be transferred between two computers. No p2p network (nor any centralized service for that matter) has enough bandwidth to handle excessive browse traffic of very large indexes without limitations. The more people on the network the worse it comes. So the real question is how to manage the inevitable limitations in a manner which encourages generous sharing instead of discouraging it.

The general answer is selective - or privileged - sharing. We are already doing it and it works just fine. We sample new music and make new contacts on the open p2p networks but when things get personally more interesting we move over to ICQ, FTP and other tools that allow private or protected small-scale sharing. With these tools we send or share large indexes to each other and enjoy the luxury of picking from them, perhaps with the help of personal recommendations. With ICQ, for example, we can receive our goods as dedicated full-album packages and even keep our trusted friends on auto-accept so that they are free to send us whatever they wish whenever they wish. These two levels of sharing - open p2p and private sharing - coexist with each other just fine, actually they feed each other. The only drawback is that we can't yet do it in the same network with a single software; we are not yet there.

Our social environment in WinMX is just a flat hotlist with no personal memories or preferences; Morpheus does not have even that. It is not hard to fathom a number of ways to make the p2p environment socially more intelligent and selective - allowing full-scale sharing for those that we prefer - but the commercial developers do not seem to show much interest in this direction of development. They either miss the point or consider it less important - or even harmful - for their commercial agendas. They see 'their' communities as numbers of users and terabytes of shared 'content' to connect and to control. The idea of private network layers and private meeting places may sound commercially risky and rightly so. The more we get connected to each other, the more we find interesting fellow peers with interesting musical tastes, the less we will care about any commercial messages. We will simply take our recommendations from those that we trust and know.

====

Mazer:
To that end there needs to be a few features in a p2p client's interface that not only facilitates but also promotes communication between users.

Ditto.

Mazer:
A p2p program that allows library browsing should have a chat screen integrated into the hotlist window. When you hotlist a user, you connect to that user's personal chat channel and can talk to that user and anyone else who is connected to that channel. Naturally people will hotlist users with the largest libraries (who also happen to have the most bandwidth) and people will hotlist libraries that have music they like. As a result people who like the same music will easily find each other.

We have to be careful though not to limit our creative imagination to the kinds of tools that Nappy provided for socializing. Nappy-like chatrooms, hotlists, IM and browsing all represent some fundamentally valuable ideas of communication within a p2p community but they must be seen as very early implementations of these ideas.

Mazer:
When a couple dozen users are connected to one user they can search and hotlist each other's libraries. The network would be small and the total number of files low, but everyone on the network would share the same musical interests so finding good music would not be difficult. This 'pocket network' would be the foundation of relationships that would last and a community that stays together. Eventually the pocket networks would merge as they find each other.

Right, if the software allowed it, it would be quite natural to see higher levels of sub-community (or 'pocket network') interaction emerging when the sub-communities would get more organized and start to offer special services to outsiders as a collective. However, the crucial thing is not to see this as an exclusive option to the Nappy-like global connectivity.

A wonderful thing about Napster was that it allowed instantaneous planet-wide adventures to all genres and to all sorts of libraries. Globality is sweet and thrilling; it is a great source of vitality, variety and interest for the entire network. We don't want to give that up on p2p. We just want to have the additional tools of social self-organization to build on the endless sea of possibilities and inspiration provided by the planetwide communion and communication with our fellow peers...

- tg