[AFRINIC-rpd] IPv4 Address Allocation and Assignment proposal

[JP Viljoen]

On 22 Jan 2013, at 5:45 PM, sm+afrinic at elandsys.com wrote:
> Hello,
> 
> This proposal is about IPv4 address allocations and assignments in the AfriNIC service region.  It incorporates the experience gained in allocating IPv4 addresses since 2006.

<snip>

Going over Andrew's response from a little bit earlier, I find that he has already highlighted a very large amount of problems with this proposal. I'd like to point out some more, and/or different viewpoints on it.

First off, I would like to state that I think this policy would give an intention for AfriNIC to move from the position of registry to the position of IP Police. A number of the clauses (as quick examples, the entireity of 9, much of 8) seem to be worded in such a fashion that it would allow AfriNIC to say whether your use of IP is - to steal some wording from Hemmingway - honest and true. This is not (and should not be) the function/role of AfriNIC, and you can verify this by looking at the roles that other RIRs play. AfriNIC's role is as one-true-source of who-has-what. Nothing more. The policies that govern its allocation practices should be ones that, as stated early in the document, benefit everyone involved. I do not believe this policy proposal benefits anyone *but* AfriNIC, and that's only if AfriNIC could maintain it successfully. Which leads me to the next point:

It is already widely known that dealing with AfriNIC is an arduous and length process, due to various (seemingly senseless) hoops one has to jump through just for the purpose of getting your network onto the internet. The reasons for this, as best as people can make out from the outside, are also reasonably well known: AfriNIC is already overloaded. This proposal would add *even more* burden to that workload, and merely on those grounds can be considered imbecillic.

Now, with the general sentiment set, let's get onto some of the particulars listed.

"""8.4 Re-numbering

Valid assignments can be replaced with the same number of IPv4 addresses if
the original assignment criteria are are still met.  The IPv4 addresses
to be replaced must still be in use.  When a renumbering request exceeds the
LIR's sub-allocation window, the request should be sent to AfriNIC for
approval.

A LIR will be given a period of up to three months to migrate to the new IPv4
address space.  The LIR may request additional time if there is appropriate
justification.  Once a network has been renumbered, AfriNIC will remove the
old assignment from the AfriNIC Whois."""

The wording in this paragraph belies the fact that the author(s) of this policy have never been involved in any form of meaningful internet-scale network operations *IN THEIR LIVES*. The very idea of forcing an entity to renumber for a new allocation is utter and complete madness. Consider, for every single growth tier an entity reaches, they would need to redo their entire network, end to end. Again. And again. This could even happen multiple times a year; for proof, see the growth of many wifi WISPs that discovered a sudden boom when they tapped into market demand. Further to this, I'm also aware of some fairly braindead pieces of hardware that do not actually even allow a change of IP, due to licensing construction and so. These pieces of hardware, or even just the license itself, typically tends to number in the multiple zeroes (of USD) when it comes to pricing, so this would be an economically unfeasible practice to do even just once, nevermind multiple times.

The general tone of the assignment procedures is also very unclear about many parts, and appears to keep revolving around LIRs becoming the centerpoint for everything. This, again, is a very bad idea. PI space, for example, is a perfectly plausible thing for an individual to have used and announced on the internet by ways of friendly networks. BGPMon, for example, gets a free BGP feed from a fairly large number of networks, and started out as a technical project venture. This is exactly the sort of thing that could get born out of PI space, by the mere result of an admin's tinkering on network stats. To centralize everything around LIRs strikes me as a "father knows best" approach, and the fact that much of the rest of this policy document makes LIRs accountable to AfriNIC for the actions of both themselves and their sub-allocations is misguided in the extreme.

In summary, I am of the opinion that this proposal would add layers and layers of complexity to daily operations (not even to mention affecting the core operations of any number of entities at present), at benefit to only AfriNIC and no other involved party.

Consider it, as it stands right now, strongly opposed.

Regards
JP