Folks,

>From the conversation so far, I glean that there are two competing PITR 
>deployment models. These are:

- User pays
- Altruistic party pays

Recent experience tells us that only the "user pays" model is viable. So, we 
are left to choose between the following variations of the "user pays" model:

1) Owner of a LISP site pays
2) Owner of a non-LISP site that access LISP sites pays

In http://www.ietf.org/mail-archive/web/lisp/current/msg04955.html, Joel 
describes a scenario that realizes model #1. In this scenario, a portion of the 
LISP address block is assigned to an EID allocator. For the purpose of example, 
assume that a /40 is assigned to the EID Allocator. From this /40, the EID 
allocator assigns smaller address blocks to its customers. The EID operator 
also operates one or more PITRs and advertises the /40, as a single aggregate 
prefix, from each PITR. The owner of the LISP site pays the EID allocator for 
these services.

In http://www.ietf.org/mail-archive/web/lisp/current/msg04965.html, Paul 
describes a scenario that realizes model #2. In this scenario, an ISP deploys 
one or more PITRs and announces the entire LISP address block (/32) from each 
of them. The ISP announces the /32 to its customers only and with the no-export 
community set. The ISP distributes the cost of the PITRs among its customers in 
some equitable manner.

The major advantage of model #1 is that it allows LISP sites to be globally 
accessible, even when some ISPs do not deploy PITRs. However, the major 
disadvantage of model #1 is that it leads to de-aggregation of the LISP address 
block. De-aggregation may be significant if customers are not required to 
renumber when the move from one EID allocator to another.

The major advantage of model #2 is that it allows the LISP address block to be 
advertised as a single, aggregate prefix. However, in this model, LISP sites 
are not accessible from autonomous systems that do not deploy PITRs.

Does everybody agree with this much?

                                                   Ron




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

Reply via email to