[AfriNIC-rpd] Re: Proposal: Reclamation of allocated but unrouted IPv4 addresses.

[Andrew Alston]

Hi Jackson



On 2011/02/09 7:41 PM, "Jackson Muthili" <jacksonmuthi at gmail.com> wrote:

> Andrew,
> 
> On Wed, Feb 9, 2011 at 7:21 PM, Andrew Alston <aa at tenet.ac.za> wrote:
>> 
>> I have concerns about this policy, since as has been stated in various other
>> discussion forums, there are several reasons to have so called "live"
>> (non-rfc1918) space that is not announced in the routing tables but is
>> actively in use.
> 
> Correct. IXPs are one exception. I am willing to accommodate other
> similar cases in the proposal.
> 
I can think of a LOT of exceptions.  Take for example the following (and I
admit this is a very specific example, but it highlights the issue)

An ISP is running an IRC server, the server has two interfaces, one globally
routable for uplinking to the IRC network in question.  One that is ONLY
announced to national transits that is the "client" interface and that is
actually visible.  This is done to avoid any form of denial of service
attack against a commonly attacked service.  The IP space assigned to that
"client" segment is DELIBERATELY not announced globally and never will be.

To accommodate this type of behavior on a case by case basis is going to be
very difficult, and to attempt to do this narrowly to specific applications
would amount to dictating what an ISP can do with their space with regards
to their routing.  The  policy would need to be modified to state that
limited announcement within a geographic area is permissible irrespective of
the reason for it, as it is not up to the RIR to dictate those cases where
it is or is not allowed to their client base.

>> Also, at which point are you evaluating the routing tables?  I can point to
>> several instances where space is "partially" announced (within a geographic
>> area, yet not propagated globally).  The space is completely valid and being
>> utilized, but factors preclude its global announcement.
> 
> That is why time from issue to announcement has been defined.
> 
The timing is not at issue there, there are many cases where this is a
permanent situation.

>> This proposal also makes no provision for the handling of so called legacy
>> address segments, which would have to be dealt with as a separate issue.
> 
> It handles legacy IPv4.
> 2.1 - 2.4 clauses cover for these scenario.
> Those are in fact the key targets for the proposal.

I have a major problem with this.  Legacy allocations were issued before the
RIR's were ever created, and were not bound by the policies that govern the
current RIR's.  As such, while the RIR's do control such services as the
whois, I believe it would be extremely problematic to attempt to force
impose policies on the holders such space.  This has been the subject of
much discussion recently on the nanog lists as well.

> 
> Cheers
> Jack