[ppml] 2002-2: Experimental Internet Resource Allocations

[Thomas Narten]

Hi Paul.

> The problem from APNIC's point of view is that APNIC members or others in
> this community may have had needs for address space for experimental
> purposes, and yet our policies made no accomodation for this.

One of the things that came out during the ARIN discussion is that
there is lack of understanding what the term "experiment" actually
means. In some cases, it is believed to be an exception to the normal
RIR allocation rules, where someone needs unicast address space, but
doesn't meet the current criteria. Here the intention may well be to
use the address space for normal purposes. That is, the "experiment"
is not an experiment in the sense of using technology differently
(whether existing technology or something non-standards based) from
what is currently in use. The issue is finding a way to allocate
address space to be used for normal purposes. In general, I don't have
an issue with this kind of "experiment".

But there are other types of experiments too. For technology
experiments, where the experimentation may potentially have
interesting effects on the internet itself (e.g., operationally), I
really think those experiments need to be reviewed carefully first by
a broad section of the community. I have the general concern that if
each of the RIRs allow such experiments, then folks will go venue
shopping to find a place where they can get an allocation with the
fewest questions asked. And for those that worry about these things,
there would now be many more places/venues where they need to be
paying attention for such proposals in order to review them all. I
don't see this as a benefit.

> A suggestion was made that APNIC implement a simple policy
> recognising this need and providing resources specifically for the
> purpose, and specifically a for limited time basis. This policy was
> proposed at an APNIC meeting, and at no stage did anyone suggest
> that the policy was inappropriate or outside APNIC's scope.

I wasn't aware of this policy being adopted in APNIC until after it
had already been adopted (and I believe others had the same
problem). So yes, you can say no one objected, but I'll argue that
perhaps the policy didn't receive sufficent review from the broader
community. I worry that a similar situation will arise when someone
proposes a specific experiment.

> > I can certainly appreciate the impetus to move beyond
> > ad hoc mechanisms.  But the proposal (this version &
> > Geoff's posted revision) addresses both points above.
> > Some explanation as to why the IETF involvement should
> > be deprecated would be helpful.

> In APNIC's case, it was never the intention that IETF's involvement
> be deprecated.  In fact section 6.1 of the doc above says
> specifically that an allocation can be made where the requirement is
> described in an experimental RFC.

In the cases where an experimental RFC makes the allocation, the
request would go through IANA. Why do we need multiple ways of
processing such proposals? The current proposal expands the number of
ways persons can make requests. I don't see that as a benefit.

>       6.2	Alternative publication approved by APNIC

> 	Experiments may be eligible for an allocation if they are
> 	described in a document that is available free of charge and
> 	publicly accessible in a forum approved by APNIC.
> 	
> 	Under this criterion, APNIC has the sole discretion to
> 	determine whether such an experiment is eligible. To do so,
> 	APNIC may liaise with IETF working groups, other standards
> 	bodies, RIRs, or Internet experts to evaluate the status of
> 	the document, the validity of the experiment it describes,
> 	and the Internet resource requirements of the experiment. 
> 	
> 	The requestors must specifically refer to the published
> 	document, describe their participation in the experiment, and
> 	provide a summary of the experiment which details their
> 	requirement for Internet resources.

Having multiple places where experiments can be approved, runs the
risk of having different criteria in different venues, which can lead
to venue shopping. This doesn't seem like a feature to me. 

> > 
> > 2/  Context -- are there specific cases histories
> > that have been problematic (for the RIRs, therefore prompting
> > this policy proposal)?  

> Proposed experimental uses have entered into the APNIC request process in
> the past, and have been dealt with in accordance with the circumstances.
> The proponents of the policy in this case anticipated more such requests,
> and although we have not made an experimental allocation yet, I believe that
> a couple of requests may be in the pipeline (not yet submitted to APNIC).

It would be really helpful, I think, if folks would point to specific
proposals and/or allocations so we can see in concrete terms what
types of experiments people are thinking about. I suspect that a large
part of the problem is differing ideas of what "experimental
allocations" are intended to cover.


> > 4/  What is "an experiment"?
> > 
> > a) A measurement effort using deployed standards-based protocols?
> >   (I.e., an effort requesting experimental allocations because
> >   the participants will not have the normal channels to obtain
> >   IP addresses/ASNs, or for which it is advisable/necessary
> >   to have all addresses from the same block, etc).
> > 
> > b) A testbed for new technology (i.e., not standards)?  (E.g.,
> >   the 6bone).

> Surely either would be candidates.

And what about proposals that were first brought to the IETF and
didn't get support there because of technical problems associated with
the proposal? The same proposal is then brought to the individual RIRs
to see if they get a more sympathetic ear there? This is an area where
I do have a real concern.

Thomas