Re: [geni-dev] CF Requirements: 2) Identity Vocabulary

On 20/03/2009, at 4:22 AM, Robert P Ricci wrote:
>
> One thing you might try is looking at the working RSpec draft from
> ProtoGENI, which is less ambitions that the spec you are looking at,  
> but
> is in better shape to see actual use.

Can you please send me a link for that?

> I think the main value in giving the computer it own identifier (as
> opposed to simply being an aggregate of the identifiers of the  
> resources
> that comprise it), besides the obvious convenience factor, is that one
> of the key things that these identifiers will be used in, for example,
> tickets, and it seems pretty clear that if I have a slice of this
> computer, and the disk gets replaced, I still want a persistent
> identifier with which to talk about the computer that my slice resides
> on.

Sure, but our main objective is to support and capture the entire  
experiment life cycle. In this case it is important that I can find  
all my experiments which used a particular failing component (or a  
newly discovered bug in a specific device driver).

The main point I wanted to make is the importance of being able to  
describe all these relationships. Obviously, not all of them are of  
interest for a specific purpose.

> We've been talking about doing something like this with human readable
> names (HRNs) that look something like DNS names. (I believe the
> PlanetLab cluster is doing something similar as well.) We definitely
> think that set of roots will be small enough for a simple centralized
> registry of them to handle. We're not using DNS and IPv6 for lookups,
> though. :)

We are experimenting with XMPP where names are somewhat similar to  
email addresses. All objects in a  domain are hosted by a server which  
you can find through DNS. The pragmatic middle ground.

Cheers,
-max

_______________________________________________
control-wg mailing list
control-wg@...
http://lists.geni.net/mailman/listinfo/control-wg