On 2/17/16, 4:46 AM, "Roger Jørgensen" <rog...@gmail.com> wrote:
[Responding to Roger's e-mail..but the answer is not specifically directed
at him.]
>On Tue, Feb 16, 2016 at 11:44 PM, Alvaro Retana (aretana)
><aret...@cisco.com> wrote:
>> On 2/16/16, 4:37 PM, "iesg on behalf of Joel M. Halpern"
>> <iesg-boun...@ietf.org on behalf of j...@joelhalpern.com> wrote:
>>
>> Hi!
>>
>>>To phrase the experiment judgment differently, either after tree years
>>>there will be sufficient demonstrated value to justify a permanent
>>>allocation, or there won't. It would take a strange situation to extend
>>>the experimental allocation (although of course we can not foresee every
>>>possible situation.)
>>>
>>>Since I do not expect the IESG to commit to specific criteria (other
>>>than those already documented in RFCs) for granting the permanent
>>>allocation, I don't see much that can be said.
>>>
>>>If you really want, I suppose that we could add a sentence saying that
>>>after the experiment, permanent allocation will be evaluated using the
>>>usual criteria for such requests.
>>
>> The point I'm trying to make is about the evaluation of what you call
>> "sufficient demonstrated value". As you say, the allocation is
>>justified
>> if value is demonstrated, how is that value demonstrated?
>>
>> At this point in time the allocation is being made temporarily so that
>>an
>> experiment can be run. What is the success criteria for that
>>experiment?
>
>I understand what you ask for, but I have no idea on how to formulate such
>criteria.
>
>What I am personally quite sure about is that the amount of
>assignment made from this allocation alone will not be a justification for
>granting this a permanent allocation. We could end up with very few
>assigment but the technical gain is significant.
Agreed.
>So, this experiment end in 3years time automatically and before that
>LISP-WG will either 1.) let it expire, 2.) ask for an extension or
>3.) ask for a permanent allocation.
>
>For option 2 or 3 to happen someone will have to present a draft that
>can justify it technical, or option 1 will happen. Either way we would
>have learned something from it I hope.
I hope we learn too.
Putting the allocation issue aside for a second...
The document itself already lists "the most relevant scenarios" in Section
3. (Rationale and Intent). For the scenarios listed, what would represent
a successful experiment? What type of information would we want to
collect to determine if the scenario was in fact relevant (or not)?
For example, taking one of the scenarios (I chose the one with the
shortest description):
Avoid penalize non-LISP traffic: In certain circumstances it might
be desirable to configure a router using LISP features to
natively forward all packets that have not a destination
address in the block, hence, no lookup whatsoever is performed
and packets destined to non-LISP sites are not penalized in any
manner.
In this case I would assume people would want to look at the impact of the
lookup -- is non-LISP traffic penalized? How much? Does having a
separate router processing help? Is it significant?
Answering questions like that would identify whether, for that scenario,
having the block of addresses made a difference or not. If, for example,
the penalty was huge and using the block reduced it significantly, then it
helps the case for an extended (maybe even permanent allocation). OTOH,
if the performance is about the same, then maybe that doesn't justify
anything. The other potential outcome is that experience was gathered,
but that the results are still inconclusive and more experience is needed,
maybe justifying an extension.
I'm just making all that up -- I'm sure the authors/WG know better than me
why each of those scenarios is relevant.
The important thing about the above is that then the future IESG is not
making a decision in a vacuum about allocation. And the future lisp WG
can also support their request for extension/permanent allocation with
data from the experiment.
In the end what I would like to see the draft talk about is not the
decision for allocation itself, but about what is going to be done with
it, and (given "the most relevant scenarios") how that experience can
support someone's case in the future.
Thanks!
Alvaro.
_______________________________________________
lisp mailing list
lisp@ietf.org
https://www.ietf.org/mailman/listinfo/lisp
