Re: draft-ietf-cidrd-ownership-01.txt to a BCP

[Yakov Rekhter <[e-mail]>]

Robert,

>     I'm sorry, but subnetting did that, VLSM did that and CIDR already did
>     that.  New practices are frequently painful.
> 
> I'm sorry, but they didn't.   Each of those three changes
> (which were certainly major practical changes) took great
> pains to avoid compelling anyone to do anything at all.

Here is few sentences from RFC1817 (this RFC "represents the IAB's 
(Internet Architecture Board) evaluation of the current and near term 
implications of CIDR on organizations that use Classful routing technology."):

   "Organizations that operate as Internet Service Providers (ISPs) are 
   expected to be able to support VLSM and CIDR."

According to your assertion, CIDR "took great pains to avoid compelling
anyone to do anything at all".  So, could you explain to me how would
an ISP that does not support CIDR and VLSM today would be able to avoid
going through the efforts of changing to VLSM and CIDR ?

   It is expected that in the near future the IANA will instruct the
   Internet Registries to begin allocating IP addresses out of the
   former Class A address space (64.0.0.0 through 126.0.0.0). The
   allocated blocks are going to be of variable size (based on the
   actual sites' requirements).  Sites that will use these addresses
   will have to support CIDR-capable routing protocols. 

I guess a site that are in need of IPv4 addresses would have a choice
between (a) not getting IPv4 addresses, and (b) changing to VLSM and 
CIDR-capable routing protocols. With choice (a) the site clearly would
not have "to do anything at all", but I doubt this would be a practical option.
And choice (b) does not fit with your assertaion that it "... took great 
pains to avoid compelling anyone to do anything at all." (as clearly the
sites has to go through the efforts to deploy VLSM and CIDR-capable routing).

> If your IGP doesn't understand VLSM, you don't have to
> have variable width masks.  

That is correct, but the "don't have to have" part of your statement has
some interesting implications. For example, if you're a provider and your 
routing doesn't understand VLSM, strictly speaking you don't have to have 
variable width masks -- you may just go out of business.

Yakov.