Recently I blogged about "a generic rdf browser is not possible". It was good to put this provocative, so some people thought about the idea, too.

Jamie Pitts wrote: "Leo Sauermann explained that a generic RDF browser needs a different display definition for each RDF-schema. This makes a lot of sense. I would add that browser-style applications which truly take advantage of RDF out there will also require something along the lines of a "useage template" for each useage scenario."

Danny Ayers blogged with something very practical: "From a random snippet of RDF/XML you can still infer quite a few things - what are properties, what are classes. Barest minimum is that you know something is to be treated as a resource or as a literal. That’s infinitely more that you get with arbitrary XML alone (you may know the datatype of something thanks to a schema, but even then you won’t know what the something represents)."

(and Shelley quoted this)

That is similiar to what I do in gnowsis' browser.

What I want to add is:
If you just have the information "this is a dc:title" than this is not enough. A snippet of RDF (an RDF chunk/subgraph/CBD) to render will most of the time describe one resource. And we have been building applications for many years now that do excatly this (Address books for persons, ERP for order management, photo finder for photos :-). Compiled applications, structured and immovable.
In the web scenario we use HTML to render this information and in the intranet scenario, many applications that were before compiled and client GUI are now intranet applications.

Building applications for certain purposes is a good thing and we should continue this tradition. One UI for photos, one for People, etc. The apple address book is loved by Mac-thusiasts, let them keep it. But RDF-ify it.

So I would continue writing web applications and normal C++ GUI applications BUT use RDF as data source and rendering source. "browsing" can then happen by clicking a "related" button and opening other applications. Same with the web browser.

Generic RDF rendering (as described by danny in short) is useful for NERDS but not for end users. I showed my generic rendering to test users and they couldn't make anything out of it. Well, probably I am a bad programmer ;-]

I don't want to build a generic "haystack" like architecture to force every hacker to code her/his UI in XSLT/haystack/XYZ. Keep your code, but make it RDF compatible.

so thats why I fumble around with gnowsis all the time, to see how application integration could work.

btw: 2 hours to alpha release.