Brendon Wilson: P2P Search Engine
Brendon had some ideas for a P2P search, where people's actual results, how they rank pages when looking for something specific, affect the results.
Users know what they're looking for and, like the US Supreme Court on pornography, they'll know it when they see it. Why not provide a mechanism for users to "close the loop" and provide direct feedback to the search engine, thus allowing other users to benefit from this extra input into the system?
It piqued my interest, and inspired this potential view of how to implement it.
What if each person, in using the search engine, could save the links that they found? This would be somewhat like a personal, online bookmark repository.
(this is for a website, but it could just as easily also have a desktop client component as well/instead)
A walk-through of a search: a user would go to the search engine. You sign up for an account, you are kept logged in with cookies, so you don't have to login every time. You input your search, and x results are returned. Much like with Google Images, clicking on a link will open the page in a new frame. You can click through links in the other frame (essentially follow a trail of links) and when you find what you are looking for, you click an "I found it" button in the P2P search frame. For your search phrase, say "ottawa baseball", you now have the Ottawa Silverblacks page associated with it.
Of course, you could then also do things like add other links or "related info" kind of pages. Not sure what the UI would be like exactly, but it would be great if the power was there to do cross referencing and see also, etc. These types of links could display separately/have different weightings in search results.
And, of course, the P2P part comes in because everyone's personal search results are weighted into the mix. An interesting add on would be to link to friends/enemies, and be able to either have their searches again be displayed as part of your own search results (Brendon's search answers for "ottawa baseball" include the Ottawa Lynx page) or be given a higher rating.
Comments
P2P Would Imply
That you'd store your data on your own device... and that there would'nt be network based bookmarks or data.... doesn't it?
Brendon has HTML up
In the context that Brendon is using it, his search is definitely P2P. He has the description up in HTML.
Which means it is a bit like my Agent Peer Search. When I had that idea, it was directed more towards files, etc. a la Napster, Gnutella, etc.
Not really
Of course, the term "P2P" is exceedingly over-used. In this case, it is "peers" that each have their own search results, and they are connecting to each others' results to benefit from combined searches.
The fact that each individual peer is a separate piece of data/chunk of database on a centralized server still makes it peer to peer.
Of course, you COULD keep those "bookmarks" on your own site somewhere, and just have a common, specified format that would be parsed and updated by a centralized server.
And hey, it was Brendon's idea -- I wouldn't have called it P2P (and we still don't know what he means, since I can't read the Word DOC).
Not in this case
The "P2P" connection in Boris' interpretation of the idea is far more tenuous than in my original idea (his is analogous to calling SETI@Home a P2P application, even though the clients never interact). In the original idea, I was thinking that the actual storage of the search index would be distributed across the network. Performing a search would involve sending a request across the network to various peers and amalgamating the responses. See the original article for more articulation.
I like the "enemies/friends" feature that Boris incorporated in his interpretation - it might be the way to lock out some of the potential for an automated attack on the system (everyone marks the attacker as an enemy, and thus his results are never featured or propagated). However, despite being on the right track, I don't think it's enough to be the whole solution.