Here at South By Southwest Interactive, the keynote interview of Daniel Ek, the 26-year-old founder of European music service Spotify, provided some satisfying insight into the major new music site.

Spotify has built a lovely music application that uses a peer-to-peer architecture to stream music at lightning speeds (a real improvement over the sometimes spotty service that comes with many other music streaming applications). The site is only licensed in Europe, but Ek says the site has 7 million users in 6 countries, and he's been working hard to get it licensed in the United States. Users can listen to music for free, with ads, or can pay for a subscription that grants access to perks such as the Spotify mobile app, and song downloads.

I couldn't help noticing, however, Ek's artful dodge to the question of how artists are paid by his service. The subject was broached by an audience member, who identified himself as an independent musician and thanked Ek profusely for the great application. He wanted to know how much he would be paid.

"It's complicated," was, in essence, Ek's reply. But he did reveal that it's a revenue sharing model; artists get paid a proportion of whatever Spotify gets paid, presumably based on the number of plays on the site they receive.

Ek's reply was disappointing because this is the million dollar question for many music sites. Pandora's been on the verge of going under for years in part because they've paid artists even when they couldn't afford to. It's clever of Spotify to find a way to be cash-positive where other sites have failed, but it means the artists must wait to be paid a fair rate.

There see other problems too. For example, pop stars are likely to draw the highest proportion of plays, but how does that relate to which fans pay a subscription fee? It seems that part of what Spotify will need to figure out is what brings money to the site and how to reimburse artists fairly.

I've been worrying about the fate of Lala ever since it was acquired by Apple last week. The speculation I've read seems split between thinking that Apple intends to embrace the company's long-term vision, creating a powerful Web-based version of iTunes, and suggestions that Apple only bought the service to poke Googlein the eye.

I first discovered Lala months ago, thanks to a deal it struck with Google, which put the service at the top of music-related search results.

When you create an account and log in, you can listen to any song in full once for free. If you want to listen to it again, you can either buy a physical CD, which also grants you permission to stream the song online, download the mp3, or pay 10 cents to buy a "web song". The web song lets you the stream the song as much as you want, from anywhere.

Web songs are exactly how I want to listen to music. I don't listen while I'm walking or commuting, but I do listen while I'm at a computer, and I want a synced service that gives me access to my songs no matter where I am. I'm happy to pay for this, and 10 cents per song is a great example of micropayments at their best--each song feels cheap, and I find I want to buy a lot of them.

Since music formats do change, what I'd really like to do is buy the rights to a song for life and have a company store it for me. But it's been hard to trust even established companies to make music available over an extended period of time. For example, when Microsoft's MSN Music store died last year the company's plans to stop running the licensing servers that authorized users to play the DRM-protected songs proved highly controversial.

For now, I'm left holding my breath over the fate of this excellent music service.

A music search engine that uses a novel technique to classify songs,will go into beta this week.

I wrote about the system a few months ago. It was designed by researchers from the University of California, San Diego, including assistant professor Gert Lanckriet. The researchers have trained the search using information contributed by Facebook users, via an application called HerdIt. The goal is to train the system to tag songs automatically--using statistical analysis applied to the waveform patterns that represent each song:

About 90 percent of the time, Lanckriet says, the system identifies patterns that are ordinarily hidden. For example, the patterns that identify a hip-hop song might include a typical hip-hop beat, but also elements that the listener wouldn't recognize as a pattern within the song. "On average, these automatic tags predict other humans' [tags] pretty much as accurately as a given human person can do," Lanckriet says.[...] He envisions a system that could take an unfamiliar song--from an independent band, or even something recorded in a user's garage--and then analyze it on the fly and suggest appropriate tags and similar music.

I'm looking forward to trying it out. See the video below for a more detailed explanation of the project.