On Thu, Feb 23, 2017 at 4:33 PM, Fernando Gont
<fgont(_at_)si6networks(_dot_)com>
wrote:
My understanding is that Randy et al are trying to get rfc4291bis to
reflect operational reality, but you want to progress the document with
no changes, essentially meaning that you want to publish a document as
full standard which doesn't agree with how the protocol is being deployed.
I think we all agree that the document doesn't agree with what Randy et al
deployed, because the document matches the existing standard and Randy et
al chose to violate the standard. What I'm saying is:
1. The "operational reality" that Randy et al want to match is not
widely deployed (in terms of number of links). The overwhelming majority of
links on the Internet (of which backbone links are just a small number)
follow the standards.
2. There are hosts that follow the standards and only support /64 in
some cases, or rely on /64 to provide useful functionality. Changing the
standards to match said limited operational reality will render those hosts
noncompliant.
3. We should not render compliant hosts noncompliant because a minority
of links purposely violated the standards.
If, even at the time of publication our documents already do not reflect
reality, we are not going to be taken seriously.
But they do reflect reality. If you look at the whole Internet, think there
are probably 1000 /64 links for every /65-126 link deployed today. Removing
the fixed /64 boundary will cause way more than 0.1% of hosts - which
*correctly* implemented a fixed /64 IID - to become noncompliant. Thus,
changing the document as suggested by Randy et al will actually cause the
document to reflect reality less than it does today.
Repeating a "classful addressing is bad" mantra isn't going to change those
facts.