new URI schemes, where to keep

In my very humble opinion, once fully understood that URIs don't have 
to be resolvable to actual, network retrievable objects, the main problem 
seems only one: unique and realiable retrival of the specifications. 

Where are the specification kepts? the RFCs are very difficult to get 
into (for both sound common sense reasons and political ones as far as 
i understand) and internet drafts are forced to fade  and deleted ( imho 
absurdely, everything should be kept for reference, that's really the 
basics, think CVS ) 

So say you have a URI scheme for some particular domain what do you 
do? you hold it up in a web site that you own, rely on its availability, 
publish it as draft .. fight to try to make it a RFC.. then eventually you're 
going to give up but what happens to those that decided to use your 
URI in their semantic web software? All they can do is hope that you 
resubmit the Draft and make noise every 6 months..  or something like 
that.

I believe that every uri scheme should be kept for reference in some 
standardized container. Then it is up to the general consensus and 
practice to decide to use it or less. Sandro Hawke came up with an idea 
for identifying generic ideas of objects or even persona moods that's 
called "taguri" (www.taguri.org) .. the funny part is that i currently 
believe in the utility of such a scheme more than he does anymore (last 
time we talked he seemed to have been convinced that is a "broken 
web architecture".. ) :-) I am writing a sw application and i believe i will 
use his scheme. 

But what will it happen to it once it has faded as draft? Should i include 
the specifications with the release of my software?

Clealy.. consensus alone cannote be trusted, the "system" should 
be stronger than that and ensure that rdf documents are readable 
and understandalbe well beyond the scope of a single organization 
or individual fanning a certain cause.  

In a not so unrelated matter..
..basically the same observations could be applied to the general 
idea of namespaces as in "location on the web where a document 
is kept".  where even additional problem pose (has the document 
been altered since the rdf was originally written? how can i be 
sure "good" on that document rappresent what was originally 
supposed to mean?) so i'd feel better with a URI for namespaces 
to identiy the "concept" and a date (concepts ARE subject to 
mutation, unless the're cristallized as RFCs are..) and of course 
many repositories working as archives capable of resolving a 
given namespace uri and give you the correct document.

--
Giovanni Tummarello