[ppml] Address space allocations for experimental purposes

[Ted Hardie]

This policy proposal has two areas which I believe require
clarification.  The first relates to the use of Experimental RFCs as a
mechanism for publishing information on the proposed use of a prefix
allocated to an experiment.  While I understand and support ARIN's
desire to have an open statement on the use of an allocation, I do not
believe this should be the primary method for most allocations.
Discussions with one of the draft policy's authors indicate to me that
there is an underlying assumption that experimental allocations
consume address space in a way different from a normal allocation.  I
think that assumption needs to be reconsidered and possibly made more
explicit.

Under normal policies, if a prefix is allocated to an organization,
that organization uses it for a period of time and may return it,
trade it in, or do any number of other things which do not require
anything like peer review.  They do require meeting the RIR's
policies.  In the normal case, that means proving need and paying
money.  I agree that experiments need to prove need, but proving the
need for an allocation seems onerous under this policy.  It will take,
at the very least, many months and the time and effort of people not
normally involved in allocation decisions.

For experiments which do not:
 
1) permanently remove a prefix from the available pool
 
2) change the way operators or implementors must treat the allocated
   prefix

3) force a global change to filtering mechanisms

It seems to place an un-needed burden on the experimenter, the RFC
editor, and the relevant ADs.

To use a concrete example, if CAIDA had needed to ask for a prefix for
its backscatter analysis 
(http://www.caida.org/outreach/papers/2001/BackScatter/), would this
method have made sense?

I suggest that the policy should indicate that allocations will be
made only when open statements are made which prove the need for the
allocation by describing the experiment.  These can be reviewed along
the lines of the "designated expert" rules used for IANA allocations
(see RFC 2434).  I suspect that for ARIN the advisory council could
serve as or select the appropriate designated expert.  Publication of
the open statements leading to short-term allocations could be along
the lines of the IETF's IPR statements; that is, documented at a well
known place on the appropriate RIR's web site.  This would work for
any allocation which, in essence, behaved like a normal allocation to
an ISP or other RIR member organization.  Under the current proposal,
an Experimental RFC is strongly preferred even for allocations of this
type.  That seems to strong and too slow.

On the other hand, where the experiment would require an allocation
that met one of the three tests above, an approved experimental or
standards track RFC does seem appropriate and should be absolutely
required.  In each of those cases, a very broad section of the
Internet community is affected on a long-term basis.  Using the IETF's
mechanisms for garnering community review and determining consensus
seems appropriate here.  I suspect it would be needed very rarely (as
most of the folks I know running experiments of this type are
deliberately trying to avoid this kind of effect), but it should be
available when needed.  The determination that it will be needed could
either be made by the experimenter directly or after the determination
that it met one of the tests had been made by the designated
expert(s).

On a related matter, this document contains the following language
on an IETF liaison requirement:

"Organisations such as the IETF, who describe experimental activities
as part of their standards development process, need to consider the
associated Numbering Resource requirements with any proposed
experiment, and under this proposal will need to liaise with the RIRs
as part of the process of publishing a draft as an Experimental RFC".

This is not clear enough to me to know how this liaison works.  As I
personally understand it, these issues are currently handled by the
publication of drafts and/or RFCs with IANA considerations sections
noting the relevant details.  This policy requests(requires?) the IETF
to take a more active role in the liaison.  Leaving aside the question
of which body in the IETF would do the work, it's not clear who they
talk to and how, or what role each plays.  It would be far easier to
ensure that the right things happen if it were more concrete, as for
example:

"When the IETF's standards development process requires a change in
the use of Numbering Resources on an experimental or permanent basis,
the IESG should notify the RIRs by electronic mail to
email at example.net on or before the beginning of the last call period
of the document which describes the change, so that the RIRs can
consult with the IETF on the use of Numbering Resources required by
the change.  Should the proposal to make a change not be in a document
which will go through last call, a message documenting the proposal
should be sent to the same address at least one month before the IESG
considers the proposal for publication as an RFC.  The RIRs shall
singly or as a body respond to such notification by mail to one of the
following: iesg at ietf.org or ietf at ietf.org."

This may well not be the right process, of course, but I believe some
language is necessary which documents the liaison process in enough
detail to allow an observer to know whether or not it was followed.  I
understand that there have been some discussions on how the IETF can
increase its cooperation with the RIRs, and it may be that this
language cannot be too precise until those discussions have been
completed; in that case, some forward reference to the outcome of
those discussions may be the best way to craft the policy at this
time.
			best regards,
					Ted Hardie