I disagree with this -- i was a very heavy Napster user in 99 and am a very heavy Spotify user (and, NB, they are a customer) today. Obviously it's a fun angle for a SXSW talk, but: - The "depth of catalog" is actually incorrect -- Spotify has far far more songs than Napster did (you can see in that screenshot even, most logins/servers hit less than 1m songs totally available, and of course those are just total index…
The backend servers were linked. If your request couldn't be fulfilled on one, your search was forwarded to the next server. The total number of files on Napster at its peek was over half a billion files. Further, that picture must have been from a server that just started because the average server had significantly more users/files on it.
Further, Napster users were ripping everything in sight. There were mp3 encodings of old wax cylinders uploaded for goodness sakes.
- The discovery on Napster was non-existent. You could browse through a users' collection, the same way you can browse through playlists today.
There was an entire curated music website dedicated to music discovery that loaded into the client.
The chat and instant message system allowed people to talk about music which created a massive music-focused community. It was wildly popular.
And don't underestimate browsing. People would search for the one song they were interested in, notice who they were downloading from, browse the other user and then start pulling down their music if they noticed several songs they liked in it. They then could send a message to that user and add them to a friend's list. That was not just music discovery, but friend discovery as well.
- I am pretty sure it was a substring match
The very first versions were substring match when there were maybe 10,000 users. Later version were not and allowed basic boolean queries like term exclusion.
- Not going to get into speed, because it wasn't Napster's fault, but even back then it was far easier to get music elsewhere other than Napster if possible
There was an algorithm on Napster that did network distance biasing. Basically, if you were an AOL user, you'd first get AOL users back when doing a search. If you were an Internet 2 or even @Home user however, your speeds were epic.
* Note: I built and ran the Napster server.