Goombah .5 released

Well everyone, this is it. The Big Day. At least for me!

We have finally released Goombah version .5. The last update, version .22, was released on 12/18/03. That's a long gap. What are we, lazy?

Let me explain... The .2-series releases were for the purpose of letting us know whether people would like the concept of Goombah and would enjoy it. And, of course, to test that the features worked as they were supposed to.

But, it was inherently unscalable. Too much data had to go from user to user, and it was all going through the server. So we limited the number of downloads of the software to 500. That wasn't enough for a "critical mass" of community to be attained, though it did allow us to receive a lot of valuable feedback.

The big thing we've been working on all this time is a p2p foundation that will make Goombah hugely scalable. It hasn't been easy to get right. But we've done it. As long as it doesn't blow up in our faces. We'll see... that's the main reason it's still a .5 beta, not .9 or even 1.0.

User reviews of the .2 series based on a poll we conducted of our users were very positive. Public reviews on the tracking sites were mixed. On VersionTracker, we got a 5-star-in-every-category review. I suspect the reviewer -- NOT someone we know! -- was basing his review partly on the potential he sees in the concept. But we also got a 2-star review there, in the first couple of hours after the original launch, when a server problem turned up and it basically didn't work.

On MacUpdate we got a 2 star review where the writer complained that it "took ages to start" and gave terrible recommendations. Ow, that hurts. The startup time was a very valid criticism, which we have addressed in a number of ways since then. But regarding the recommendations... of course the recommendations sucked: collaborative filtering systems need a sizeable population of users, and we had intentionally limited that release to only 500 downloads. (And for all I know, that reviewer may have complained when there were still only 10.) The new release is inherently scalable. Gradually, the recommendations will get better and better...

In fact, the actual megalomaniacal intent is to produce the most accurate recommendations the world has ever seen. There are very specific technical reasons why I think we can actually do it. That's one of the main design goals behind Goombah. But it is a simple fact that the recommendations will get asymptotically better and better as the community grows.

The way to look at the recommendation accuracy question is as follows. For a collaborative filtering system to make good recommendations for a particular user, there need to be other users like him in the system. The more similar they are, and the more of them there are, the better the recommendations will be.

So people's whose tastes are somewhat mainstream will get good recommendations from Goombah now. People with highly unusual tastes will get steadily improving recommendations as there are more other people like them in the community. But we have to start somewhere... if all goes as we hope, we'll grow steadily over time and meet the needs of more and more people.

Goombah isn't only about recommendations. It's about seeing what other people have in their collections and how much they tend to play each track. That way you can explore new music... it's also about blogging, and communicating with other people whose music collections interest you...

So, we don't ever expect to be "all things to all people", and particularly not at the beginning. And today is really the beginning. We expect to get all sorts of positive and negative feedback as we grow. We'll take to heart all the negative feedback and do our best to address any problems. And we'll take heart from the positive feedback. We'll keep on truckin, and getting better, no matter what.

I hope that you'll download Goombah and give us your feedback.