Okay, we need to reset this thread. A little history:

(1) I originally suggested to request IANA for a well known /12 prefix that was 
known to be an EID prefix where you could be guaranteed to have xTRs deployed 
at the site.

(2) I requested this so when a packet hits an xTR at a LISP site with a 
destination address not in the /12, AND a configuration knob was set, the 
packet could be routed natively to the non-LISP destination site.

That is all I wanted out of the /12. It was a way to make LISP to non-LISP 
communication avoid a mapping database lookup that would yield an answer of 
"destination is not LISP", which the well-known prefix would tell us right away 
(yes, by hard coding it).

The thread has wandered into:

(1) How will registries allocate and how will sites get EID-prefixes. 

(2) How can LISP succeed without such coordination EID allocation.

(3) If this is experimental versus production.

Where in fact, any address today can be used as an EID-prefix by simply 
allowing an xTR to look it up (the configuration knob above is not set) to 
determine if there are xTRs at the site. This is why I keep saying there is no 
requirement for RIRs to allocate EID-prefixes.

Dino

On Mar 12, 2013, at 4:47 PM, John Curran <[email protected]> wrote:

> On Mar 12, 2013, at 7:36 PM, David Conrad <[email protected]> wrote:
> 
>> On Mar 12, 2013, at 4:21 PM, John Curran <[email protected]> wrote:
>>> The choice of going with a single registry with neither competition or nor 
>>> elected 
>>> oversight may also work, but places enormous faith in the enduring best 
>>> intentions 
>>> of the enlightened and benevolent overseers, particularly once there is 
>>> significant 
>>> usage and a complete lack of viable alternatives. 
>> 
>> I dunno -- IANA seems to have done pretty well with the 1000+ protocol 
>> parameter registries they manage with lightweight oversight by the IETF/IESG.
> 
> Indeed.  While I don't believe any of the parameter registries are as 
> widespread and pervasive as what is intended with EIDs, there is no 
> reason to believe such scale isn't achievable.  I was only noting that
> the reasons expressed on the list for having a registry/registrar split
> included more than just "scalability".
> 
> FYI,
> /John
> 
> Disclaimers: My views alone. This email sold by conceptual weight not volume.
> _______________________________________________
> lisp mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/lisp

_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp

Reply via email to