[sldev] Open Grid Identifier Notation draft proposal

[Cenji Neutra]

> Cenji Neutra wrote:
> To my mind, the identifiers aren't resource locators (URLs).  A URN does
> seem like a nice idea though (goes looking for the URI RFC...)
>
On Tue, Jun 10, 2008 at 3:29 PM, Vex Streeter <vexstreeter at gmail.com> wrote:
> XML namespaces ran into the same confusion, where the namespace identifiers
> *typically* look like URLs but are often actually unresolvable because they
> are "merely" namespace identifiers.  The RDDL idea is to make them actually
> resolvable URLs, where the intent is that they resolve to a rddl document
> that describes the domain that the namespace models. The nice thing about
> this approach is that it allows for a completely distributed solution: once
> you've got the identifier, you can find out everything else you need to know
> to connect without  any sort of central registry.  The point to make here is
> that even in the URL case, the identifier is still just a reference not the
> grid it names.
>

Yes, that is a nice property.  The only problem with it is that it
requires that everyone who controls the domains corresponding to the
identifiers to implement the standard service.  That would be like
requiring hosts on the internet to run their own DNS name server for
the subdomain below (if any).  While they can certainly do that with
the current DNS delegation mechanism, they're not required to do so -
they can have an independent DNS service provider (who has no control
over the hosts in the domain) for a whole set of hosts and subdomains
if they wish.  I hazard a guess that the current success of DNS
wouldn't have come about if it had required everyone to implement it
on many hosts from the outset.

Perhaps it is possible to have the advantages of both systems.
Perhaps we can use URIs and specifications that layer above that can
require that for lookup, applications first query an extended DNS
name.  What I mean is, if the identifier is some URI containing
"main.mygrid.net" then the application should first query a URL like
"http://main.mygrid.net/grid.xml" (or whatever) and if it receives a
404 or other error (including a timeout) it should then try
"http://main.mygrid.net.virtualworld.net/grid.xml" where
"virtualworld.net" is the configured lookup service provider for that
application (and any number of parties can opt to provide that kind of
lookup service).
That way, we can start providing a conforming service that works for
all the existing grids (and new grids) even if they haven't
implemented the standard.  The 'timeout' could be a problem.
Perhaps we should just use DNS to do a SRV record lookup of the domain
and the absence of a record would indicate the service isn't
supported, the presence of a SVN record can point to the URL of the
actual XML.  If there is no record, then a SRV record lookup on
main.mygrid.net.virtualworld.net (or whatever) is performed instead.

Thinking out loud there :)
-Cenji.