Musician Archive

/nod

This would be nice, but because of copyright issues, it is doubtful the idea could ever be implemented. There are a few other threads about this you can take a gander at for more pros and cons about this idea.

-Liloman Hammons--Novice Pistoleer

I think the other issue maybe the songs are stored locally, so all you have to get from the server is a signal that says what song you are listening too, and if and which flourish the performing. If the songs were created by the players, then you would have to upload the song to the server. This could be done at song creation time; however, then due to the copyright concerns (I know lucas arts is big on this one, our CS dept got in big trouble for serving episode 1), the song would have to be validated by the server (or farmed off to a special server to handle these requests, but due to the number of songs, this could be a major waiting issue.If this was accomplished, then the audience members would have to download the song from the server. The upload/download could be implemented, at some considerable bandwith costs. Imagine what this would do for the lag in coronet?

***warning next part is a little more technical, read on if you have some understanding of algorithms and computablity. I am feeling wordy, and just feel like trying to tackle this problem, lol. This is a first look at this problem, and I would appreciate any constructive comments pertaining to this discussion.***

However, the validation would be very difficult to implement (if not impossible, hasn't been proven yet). The first issue is to define what is not a valid song. The issue lies in a small portion of one copyrighted song makes another song invalid if it is copied. I will ignore the basic rules of music creation, such as being in the same key or modes.I will define an invalid song as one that matches another portion of a song by a small portion.

To begin, a song is composed of a string of notes each varying in duration and what the actual tone represented. All possible combinations can actually be enumerated in an alphabet. As an example a quarter note in the tone of middle C can be represented as qmc, so the next note can be represented as qmd. Now that we have an alphabet, we can define a song as a string of this alphabet. At this point I have made this a problem of substring matching.

Hopefully you can see the issues involved at this point. First we have define the whole alphabet, then we have to create a database of all copyrighted material that has been converted into this string format. That gives us alarge (but finite set of songs). Now we have to define a size of a substring that would be a copy violation. Lets say this substring consists of 10 notes. The nice thing is we can ignore strings of any other size since, they either have to match the 10 notes. Now we have to generate all 10 note substrings of a song that was created by the user. The best string matching algorithms can perform in O(n) time and O(1) space. Hence, we could have up to k substrings in a song, and a very large N songs in our database, to check against. I'd say it is reducible to a unsolvable problem, but it's bed time now.

***brief summary***

This problemof verification, is similair to, finding if a sentence you wrote is ever copied in all published works. Or, another analogy, finding a certain combination of letters such as "qtn" appear anywhere in an entire dictionary. The only way you could verify is by checking every word.