From a blog-posting I made here : http://platformwars.blogspot.com/2005/11/world-of-ends-applied-to-synweb-vs.html
Hmm. I wonder if another way of thinking of this is a world of ends type argument with the data-format playing the role of the network-protocol and programs playing the role of the ends.
The OPML format doesnt know or care whether its carrying a playlist, a blogroll, a blog, a subscription-list etc. It just concentrates on allowing the data to be moved from one program to another. Meanwhile, the OpmlEditor is continuously being upgraded to know how to get more meaning from the outlines it creates.
Why is it better for the meaning to reside at the edge?
- a) edge points can be upgraded individualistically. If I want OPML to represent my attention data I just have to upgrade to the new version of FeedDemon which will support this. If I want to upgrade to a new version of a specially crafted Attention RDF I have to get the whole network to buy-in.
- b) The corollary of that is that if youre crafting a model like Attention RDF, its really important to get it right. Because upgrading is such a pain. On the other hand, its far less important to get attention in OPML perfect from the start. It can be continuously tweaked with new releases of the software.
This again reflects on the question of applications. OPML as a format is worse than XoXo or some other XML. But its gonna be the winner here as long as Dave Winer can keep up the development momentum in the OPML editor and the rivals dont offer a compelling alternative.
LesOrchard writes a great concept demonstrator but then abandons it. [Danny] does a lot of coding. But because [he's] focused on putting the semantics into the protocol, [he doest treat his] programs as the primary vehicle for getting his meaning into the world.
If I want to use Attention RDF as a user, what do I do? Wait around until people have thrashed out the spec on the wiki and then install a very generic RDF database and query language? Why wouldnt I rather follow regular updates of an existing, already useful tool, which is promising to add this functionality in small sips over time?
 I recognise the usual respost : about atomicity of SemWeb tags which means it isnt an all-or-nothing thing. Its just that I dont see that this actually helps here.
For example, what would be an incremental roadmap for geting to widespread Attention RDF adoption?
We cant say well, well roll out the att:readtimes tag to begin with, and then once people have started using that and seen the benefits, we can add att:lastread.
But thats exactly what a program-oriented SynWeb strategy can do. Well start with rank in the next version of the code, and then go from there.
In an earlier comment, I'd said something to the effect that there was an 80/20 rule. If 80% of potential users were satisfied with a format, that would make for a viable standard.
Danny asked : There should be plenty of data points, but thinking about it theyre rather hard to pin down.
So heres an interesting one I thought of : 8-bit ASCII - where the second 128 chars are indeterminate and available to be adapted by whichever country or special interest that wants them. Theres lots of incompatibility as people try to pass files containing accented vowels from one country or operating system to another. But ASCII is a phenomenally succesful standard despite all this potential for error. Most of the time, most people, stick within the working subset.