Re: globally unique identifiers for lightpaths?

[Freek Dijkstra <fdijkstr@xxxxxxxxxxxxxx>]

Jerry,

You make a valid point about the visibility of identifiers. Why do we
need it? Typically, one domain wants a common ID for a link through its
domain to talk to its neighbours about it.

This stitching of links allows that.
I may interconnect 20071130180741-maxgigapop.net-uva.nl to
20071205200042-uva.nl-cesnet.cz, creating
20071206160626-maxgigapop.net-cesnet.cz. This is nothing different the
fact that 20071130180741-maxgigapop.net-uva.nl may be composed of a link
20060215060213-uva.nl-netherlight.net,
20051108095411-netherlight.net-manlan.net,
20071128215210-maxgigapop.net-manlan.net, and
20070707140212-maxgigapop.net-nasa.org, and that this last link is
actually composed of a set of even smaller links within your domain.

Such a hierarchy is inevitable in the end, if you forgo central
know-it-ALL control. So all we need a link ID, that is agreed upon by
the domains it goes trough, and by the upstream "user" of the link
(which may be a user or an agent providing it on behalf of a user).
Also, this link ID is know by the sublinks directly underneath.

That is why I suggested to name the link after the requester. If you
need to provide a set of smaller internal links for this big link by the
request, you may give that smaller set of internal links a name of its
own, and you tell the requester of that name, not of the individual
internal link names.


As an aside, you claimed:

>   <Global_resource_ID>  ::=  <admin_domain_id>-<admin_resource_id>
> 
> I don't think any information about the light path
> itself should be present in the GRI.

Well, in your case it does: it represent information about the admin domain.

It is not bad to have this information (in fact, it is even useful). But
it should not contain any other information.


As a clarification of my previous mail. Why do we actually care about
the format of this ID? If it is a globally unique identifier, why not
use an UUID? 39377f80-a4d5-11dc-bbdf-0002a5d5c51b as just generated by
http://www.itu.int/ITU-T/asn1/uuid.html really is unique, and does not
contain any information. Exactly your two requirements!

The only reason why a common format is useful is for software (maximum
length of a name), and human recognition (if we use the same format, it
is easier to recognize a link name).

Since our aim is to make this all automatic, there is no need for human
readable names. However, we are not there yet, and adding human readable
info to a name helps in the adaptation and transition. So that's why I
prefer a URI, 20060215060213-uva.nl or 20060215060213-NLAMS-USWAS over a
UUID.

Just to be clear: while I advocate that a generated name like
20060215060213-uva.nl or a URI over a UUID, I think that any control
plane should treat it as completely opaque. I'm aware of the security
implication: I could generate 20060215060213-keio.ac.jp for a link
between Prague and Amsterdam. But as long as the name is only used for
easy memory, not for any assumption, that is OK.


Regards,
Freek

Attachment: signature.asc
Description: OpenPGP digital signature