I think that one thing that's a bit of a problem in Audiosurf right now is that not everyone gets their tags from the same sources. For example, I tag all of my music using MusicBrainz, but apparently not everyone does. I have a lot of stuff that is in the original language, but a lot of people will have transliterations (I guess I'm going to have to go through and manually retag all of these or something, though if I can figure out how to use it I think Picard has a plugin for this that would at least help.) It's too bad now that Audiosurf has such great Unicode support, but it seems that all too few are taking advantage of it apparently. And just recently a friend posted a score on one song, but her copy had the artist's name reversed (actually technically correct, but no one uses it that way) from the ordering everyone else was using. As such, her score isn't even on the global board for that particular song, but is instead completely on its own. I did a run on that song and was quite frustrated as I had to restart several times and such due to some really tight spaces between the grays and by the time I finished there was no way I was running it again... (Yeah, we can compare the hard way, but what's the point of the global boards if people have to do this?) Most importantly, I think everyone playing the same song should be comparing scores rather than everyone playing the same tagged song...
Now Picard and such aren't perfect by any means. I've had a perfect copy straight from a CD where Picard thought most of the tracks didn't match the album for some reason (and one or two where it tried to match a track or two to another album or a single -- sometimes not even the right song.) I will say I've had better results with the MusicBrainz's algorithms than any others so far though. So I don't think it should be enforced to this or something since there are incorrect results. I think maybe it should try to check each song while it's generating the track data and if it finds a match, ask the user if this match is correct. I think it should maybe save the MusicBrainz album/track IDs in the metadata for the file, but not change the actual artist/title/etc data in case the user has them a certain way on purpose. I do think that if it were made optional it should be turned on by default as many people don't really look through the settings much and would just leave such a thing turned off, thus defeating much of the purpose here.
Now obviously this has a couple of obvious problems. First, MusicBrainz lookups aren't exactly blazing fast. This should perhaps be an optional thing and obviously the results need to be saved locally in some manner so that each particular song would only need to be looked up once. (The game is already saving data on each song anyway, so this seems to me like it wouldn't be a big deal as we're talking about likely considerably less than 1KB of plain text.) Also, MusicBrainz may not be super-happy about the volume of requests to their server. I would propose that to this end, AudioSurf could host its own proxy server for this that does some major caching. All of the popular results would be looked up instantly in the local cache and thus all of the more popular songs would result in a non-hit to MusicBrainz servers (and I imagine this would make up the majority of the volume of requests -- especially given that the worst hits would come from the likes of the radio, though I would imagine that the radio could download this information while it's downloading the song anyway.) I think the actual volume of data would be relatively small -- especially compared to what the radio is using up -- and with basic HTTP compression enabled it should really be quite minimal by comparison I would imagine. Even the cache probably wouldn't be much given that we're talking about data that consists more of small amounts of plain text than anything else (and obviously it doesn't even need to bother with loading ALL of the album data. For example, the Amazon ASIN has no value in Audiosurf and thus should be skipped.)
I realize that not everything that is being worked out with MusicBrainz is complete and, in fact, it's not being used right now, but I think that in the discussions with MusicBrainz, maybe something of this should be discussed as well?
Alternately, I'm not quite sure how it could be done, but perhaps some sort of "redirect" system could be implemented. For example, Last.fm is now starting to utilize this sort of concept. Many things that are transliterated will redirect automatically to the correct thing for example. Artist names that are reversed are often switched back or even transliterated back to the original text sometimes. As such, a lot of incorrect tags (or technically correct, but different from what the majority use) are now being redirected to the correct artist/song so that the entry will work much better. (Especially useful for the charts since if you listen to the same artist and their name is reversed on some of the songs but not all, it throws off the whole thing.) Such a system pretty much has to rely on user corrections though and this obviously isn't perfect by any means, but it still works a lot better than without it.
Just a thought anyway. Whatever the case may be, we certainly need some way of getting differently tagged same songs showing up on the same charts.
