Re: distinguishing between 'description URLs' and 'interface URLs'

On Wed, May 9, 2012 at 5:39 PM, Michiel de Jong <michiel@unhosted.org> wrote:
> The words URN and URL already exists and are clear to everybody i think.
>
>    'interface-URI' and 'description-URI' are the two terms we need to
> introduce IMO.

Actually after thinking about it for another day and talking to Melvin
some more, i think that this not an ideal classification either. I
think it might be better if we distinguish primarily between direct
and indirect URIs.

here is my understanding of how each one works:

given that X is the direct URI of Y. what can we conclude?

if X is a document URL, then Y is the document that can be retrieved at X
if X is a message queue URL, then through Y, a message can be sent to
X. Since we can't model what the agent behind that message queue looks
like (this is like the Turing Test), it's OK to model the user and the
mailbox as one amorphous entity.
otherwise (i.e. X is a URI that is a URN and/but not a URL), then Y is
whatever Y means according to its URN scheme.

So with this, we can model, describe, and reason about:
- documents
- email recipients and other recipients of other message protocols
- books (through urn:isbn:)
- users (through acct:user@host) - Melvin doesn't particularly like this one ;)
- browser settings and environment variables (through about:)

Direct URIs are what make up linked data. But how do we reason about
the rest of the world? Trees, birds, buildings? They are not on the
web and have no direct URIs. Here, indirect URIs comes into action. We
start using URIs in three extra ways:

if Y is an indirect URI of X, then:
- if Y is the URL of a document, we know that at Y we can retrieve a
document that is a description of X in some language.
- if Y is the URL of a message queue, we know that we can send a
message to Y, and if its delivery is successful, then X would be the
entity receiving it.
- if Y is the URN of a book or some other text medium, then that
medium contains a description of X in some language.

Melvin tells me that indirect URIs can only point at document
fragments, not at whole documents. i find that very arbitrary, so if
that is indeed the current standard then i think that should change.

before we interpret a URI, we should just always ask 'is this a direct
or an indirect URI?'. if the answer is not known, then our computer
program should throw an error.

also, if this is how we understand the web to work, then we should
describe it this way in our presentations and wikis. IMO, not
mentioning the difference between direct URIs and indirect URIs leads
to endless discussions and confusion.


My 2ct,
Michiel

Received on Thursday, 10 May 2012 13:44:40 UTC