Hi Noel,

I have been dedicating a little time each day to a close reading of 
draft-ietf-lisp-architecture and will generate a few comments each day as the 
work progresses. The following are comments on Section 4 (Architectural 
Aspects).

Section 4.1: Critical State
===========================
This section would be more concrete if you identified the state that you 
considered to be critical. We can probably agree that mapping information, as 
it is maintained within the mapping system, is critical. However, reading this 
section, I can't tell whether you consider the map-cache, as it is maintain in 
an xTR, also to be critical. Could you clarify this in the text?

Also, if you consider the xTR map-cache to contain critical state, you should 
probably note that staleness, as well as loss of this state can cause loss of 
connectivity.

A New Section Between 4.1 and 4.2: xTR State
============================================
You might want to add a section stating that LISP distinguishes itself from 
other routing architectures because:

- it maintains a cached-subset of mapping information on the xTR
- the xTR pulls mapping information from the mapping system to itself (as 
opposed to letting the mapping system push it)

With this architecture comes benefits and costs. One benefit is memory savings 
on the xTR. (There may be other benefits). The cost is that the protocol needs 
to address the problem of cache freshness without compromising performance or 
security. 

Section 4.3: Piggybacking of Control on User Data
=================================================
- Would it be fair to say that a major motivation for piggybacking control 
information on user data is to address the problem of xTR cache freshness? That 
having this information reduces the need to expire map-cache entries more 
quickly than might otherwise be necessary?

- Would it be fair to say that the piggybacking of control information on user 
data reduces the amount of ETR probing that is required to maintain cache 
freshness, but does not totally eliminate it. 

- Would it be fair to say that piggybacking of control information on user data 
introduces a class of security issues, many of which are addressed in 
draft-ietf-lisp-threats?

- If any of the above statements are true, it might be helpful to add them to 
Section 4.3.


--------------------------
Ron Bonica
vcard:       www.bonica.org/ron/ronbonica.vcf



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

Reply via email to