Re: [Anima] Warren Kumari's No Objection on draft-ietf-anima-prefix-management-06: (with COMMENT)

[Brian E Carpenter <[e-mail]>]

On 14/12/2017 09:48, Warren Kumari wrote:
...
> ----------------------------------------------------------------------
> COMMENT:
> ----------------------------------------------------------------------
> 
> Thank you.
> 
> I did have some comments / questions.
> I'd also like to draw both the authors, and AD's attention to Fred Bakers
> excellent thoughts in his OpsDir review -
> https://datatracker.ietf.org/doc/review-ietf-anima-prefix-management-06-opsdir-lc-baker-2017-10-23/
> 
> Firstly, a global concern:
> This technique (and I suspect many automated prefix allocations where a device
> uses space, and then requests more) is likely (I think) to result in
> fragmentation of the address space - this will lead to more routing entries in
> the IGP, which may be an issue for smaller routers or "L3 switches". I think
> that it would be useful to note this.

That devil is in the details, but you're correct, that is a risk. How big a risk
depends on the algorithms and policies used. If we get to update the draft,
this would be a good point to add.
 
> 
> I also wanted to make sure that the author of this document were aware of the
> CASM BoF from IETF98 - I've just checked, and see that at least Qiong Sun was
> associated with the work (draft-xie-ps-centralized-address-management).

Yes, in fact it would mesh quite nicely with CASM. That's the main reason
I pushed to have the C in CASM mean "coordinated" instead of "centralized".
 
> I had a question -- I don't really understand what: [Page 9] "A gateway router
> in a hierarchical network topology normally provides prefixes for routers
> within its subnet, ..." is trying to say. I've seen many "hierarchical network
> topologies" and don't believe this to be true, nor do I really understand what
> "its subnet" means. In some cases a router will announce an aggregate for
> customers behind it, but I don't really view that as a general case. I'm
> guessing I'm just not understanding - can you please educate me?

Hmm. In a manually designed network (especially with v6) I'd expect
the initial design would be done in something close to a binary tree,
but in the real world things tend to drift from that starting point
over the lifetime of the network. But I think the phrasing is probably
trying to say too much in too few words. All it's really trying to say
is that the ability to negotiate with an arbitrary peer is more powerful
than only being able to ask your upstream for more prefixes.

Again, happy to clarify the wording if we update the draft again.

   Brian