On 2011-01-29 15:49, jamie baddeley wrote:
I don't think the proposal at this point is too bad. Happy to be persuaded otherwise.
If someone is prepared to make the leap into a /22 then 20% of what's remaining in the
existing upstream allocation is not a massive amount of space. How many assignments happen
from upstream to downstream that is greater than a /22?
Yes, certainly this isn't a disaster, but why set the level at 80% occupied?
90% would halve the amount of potentially wasted or hoarded space.
We're expecting everyone who takes the global
routing table to be (or have been) busy upgrading to v4/v6 dual stack and I presume as a
consequence they have nice shiny routers with lots of mem/cpu etc. Therefore the number of
prefixes in the GRT is much less a concern these days right?
Well, routers have kept up with growth because CIDR has been a great
success over the last 15+ years, and this seems like a step back:
2 prefixes instead of one for every operator who uses this policy and
has multihomed transit. There's a significant risk of IPv4 disaggregation
for many reasons during the coming address space end game, so I think we
need to be watchful.
As an author of RFC 5887, I fully realise that asking operators
to renumber is a hard ask. So the effect of this change will actually
be to nullify the renumbering requirement completely - why would
any operator take that pain voluntarily?
On 28/01/2011, at 3:22 PM, Brian E Carpenter wrote:
The APNIC policy proposal 94 allows an operator
to put in for new
IPv4 space without having to renumber, if they can show that they've
used 80% of the space already obtained from their upstream.
That seems bad in two ways
1. It allows 19.99% of IPv4 space to be hoarded.
2. It probably encourages disaggregation, compared with
renumbering into this new block.
NZNOG mailing list