There've been a number of posts on the subject of compatibility based on shared interests. The possibility of finding some unknown other, or group, who share a large number of your recorded interests is clearly something that would be intriguing to many, but is sadly impractical given the size of the LJ directory. Clearly we need to find a way to reduce the number of users to be searched through.
I've thought of / stolen a couple of possibilities beyond the friends list (which is rather limited since I'm presuming that much of the curiousity comes from possibly meeting new friends):
- searching the membership of a given community
- searching within a given geography
- searching the friends list of a given user
- searching the subscription list of a given syndicated feed (e.g. apod)
- starting with a given interest (or combination of interests) and then searching for the best match amongst users who have that interest
- searching the last X journals to be updated
- searching the friend of friends list
I strongly suspect several of these wouldn't be feasible. Of all of them, I particularly hope that it would be possible to start with a given interest, use that to preselect a number of livejournal users and then proceed from there. For example, from my interest list, I could perhaps choose "first world war" - an interest that my interest data page tells us that I share with 39 other users (although the data about the interest itself seems to disagree, saying that 32 users and 4 communities - i.e. 36 entities - share the interest, if someone can explain that). These 39 (or 32) users could then be analysed to find that user with whom I share the most interests.
There's still a problem, in that if I were to enter 'music' as the initial filtering interest, the code would still be presented with 1,135,483 users to analyse, which although greatly less that LJ's 7,356,123 users (as of today, right now), still represents a very significant load. Any meme that proceeded along these lines would have to either reject any interest that didn't bring the numbers down to a reasonable figure, or possibly offer to cross reference the list with another chosen interest (i.e. narrow the list down to users interested in music AND art AND red wine, say).
Certain of these ideas - for example searching within a given geography / community / interest - could be extended to searching for people with the most friends in common, or people who are members of the same communities as the user, or members who follow the same feeds.
On a completely different note, I have another idea for a very simple meme that might be of interest to someone with the necessary skills: I'm sure you're all aware that June 2005 is seeing a permanent account sale. I was weak and craven and greedy, and I wondered how many among my friendslist had been similarly so. If the information is available, it strikes me that it'd be a fairly simple meme to extract account type information from a user's friends list and turn this into a chart, showing the proportion of free, paid, permanent and early adopters amongst one's friends list (and for extra interest, comparing the distribution to that of LJ as a whole, possibly including a measure of eccentricity).
Finally, I've recently started programming 'properly' (as opposed to picking up bits here and there, doing a few exercises and then running out of steam because I didn't have anything real to apply it to) - picking up Rexx / Object Oriented Rexx and then actually using it for work, using it to write some fairly straightforward mainframe scripting / check applications. This has spurred me to want to pick up something more mainstream, but it's also taught me that there's no point unless I can find something that I'll be motivated to apply what I learn to. Hence, my question: I have a website on a server that supports PHP hosting - is it worth my while trying to learn PHP for meme-type applications, or is it not really powerful enough for that?
Thanks for reading, if there's nothing worthwhile among my ideas, I'll not be bitter. This forum for collaberation is a great idea and some of the names I see attached to it have been responsible for some excellent memes. Keep up the good work!