Hi Tom,
Appreciate your comments and ideas.
Just to clarify a few things though because words can sometimes be used incorrectly or without precision of meaning.
1. Attention is not a format or an algorithm. It is the giving of focus from a human being to 'something'. Measuring and using Attention for the benefit of the user requires a number of complex moving parts up and down the stack. There is no silver bullet.
2. Therefore, the information that Last.fm and other sites let you publish via RSS etc is not simply 'Attention' - it is a specific compaonent of the stack we call "Attention Data'. A trail that indicates you have paid some Attention to something. In this case a song.
3. APML stores 'Attention Profiles'. These are very different from Attention Data. They are a boiled down result rather than all the contributing data.
4. You suggest that instead of creating an intemediatory format like APML, users should just link all their private and public feeds of Attention Data into Amazon (for the sake of argument).
This could work, however consider that:
A. This is a fairly specialized, onerous and extremely self-revealing task for most users. While they want Amazon to have an instant 'handsake' they don't neccessarily want it to know everything about them. They may want to pick and choose profiles and modes of privacy.
B. No one is better positioned to know exactly what 'Listening to a song' REALLY means than the tool you used to listen to it. So to ask Amazon to process your Attention data from Last.FM - and any number of 3rd parties - will result in less accurate results. It is better that Last.FM uses it's intimate knowledge of music to extrapolate the meaning of 'listen' for it's particular tool/user experience.
C. The values mentioned in APML are a percentage. They exist along a well defined liniar scale so they are quite reversable. However you are right that the algorithm is de-coupled from the data-store. It is deliberately open-ended. Why? To allow for innovation.
Google and MS Live Search both crawl the same web - the same data - however they each give different results. The same will be true for APML tools - and that's perfectly OK.
Vendors are building tools and it will be up to the user to choose the tools they like based on outcomes that meet their needs.
In the end, however, this is not so much a technology debate as it is a philosophical one. Users own their own Attention Data *and* their own Attention Profiles. On that, I think we both agree 100%.
Hope this helps clarify the thinking for APML.
It is not Attention Data... It is an Attention Profile.
Think of it like ZIP or RAR for all your Attention Data.
