It doesn't have to work on everything to suceed, as long as:

- There are large enough subsets to be worthwhile in their own right (e.g. J2ME phones, all Blackberry handsets, etc).

- Customers only see apps that will work on their phone. At present the mobile app industry is not helped by confusing customers before they even make a sale - not everyone knows what their device can do.

- Quirks that make porting between devices hard (e.g. wierd handset bugs, or carrier specific changes) are documented well. T-mobile are probably already doing this and it would cut down development time for ISVs.

I don't think it'll be a failure: I think they've got a good shot with this. What this does, however, is silence critics of Apple's App Store plan. When other companies pick it up, it gives Apple credibility.