Re: [Roll] WorkingGroup LastCall:draft-ietf-roll-protocols-survey-02

Hi Thomas:

Every evaluation picks criteria and applies them. This doc has been around for awhile with its own clear list of criteria, which take their source in the requirements drafts. Someone could propose another doc with another set of criteria that would make perfect sense for a different set of requirements.

We could make a moving target of this doc and pursue the quest of the perfect evaluation of life, the universe and everything. But that's not the motivation nor the best interest of this group. In any case, do we expect a perfect one-fit-it-all protocol? I guess not. We'll have to trade off and prioritize, so while the evaluation draft has mostly served its purpose, we are far from being done with the requirement drafts themselves. 

For instance, in the case of industrial, there is an emphasys on reliability vs. path cost, which *is actually* achieved in the real (non-IP) world by building graphs that tend to enable multiple forwarding solutions at each hop, with the capability to revise the forwarding decision on a frame by frame basis after a few L2 retries. 

There are deployed IP protocols that enable similar non-equal-cost path computation, for instance in the DV world with EIGRP feasible successors. There is a huge amount of reusable art in traffic engineering. There is also interesting art in academia such as Tora that might fit the bill. What we need to do is build on this experience and start the work on real solutions to figure what optimum we can actually achieve for our requirements. 

Pascal

>-----Original Message-----
>From: roll-bounces@ietf.org [mailto:roll-bounces@ietf.org] On Behalf Of Thomas Heide Clausen
>Sent: mardi 9 décembre 2008 16:21
>To: JP Vasseur (jvasseur)
>Cc: Dearlove, Christopher (UK); Emmanuel Baccelli; roll@ietf.org
>Subject: Re: [Roll] WorkingGroup LastCall:draft-ietf-roll-protocols-survey-02
>
>
>On Dec 9, 2008, at 16:13 PM, JP Vasseur wrote:
>
>>
>> On Dec 9, 2008, at 4:03 PM, Dearlove, Christopher (UK) wrote:
>>
>>>
>>>> 2- There is clearly a need for a solution *today*: look at all other
>>>> (deployed) non IP solutions out there
>>>
>>>> 4- We are not talking about a "dream protocol": again this is very
>>>> much realistic since solutions do exist today (not IP unfortunately
>>> :-( ....)
>>>
>>> Can you point me to an existing (doesn't have to be IP) protocol
>>> that passes all five of the ID's tests, without some major drawback
>>> in some other area. I for one would find it interesting and useful.
>>> (Obviously it would have to have a specification, not just a claim
>>> made about a proprietary protocol.)
>>
>> You missed my point ... I did not say that there were any of them
>> satisfying these criterion.
>> But clearly there are non IP solutions being deployed, my point
>> being that we need an
>> IP solution soon and we should make it right.
>
>I read Chris Dearlove as asking for something concrete, justifying
>that these five "criteria" that the survey I-D sets out are
>*attainable* (and without some major drawbacks on the other important
>metrics that the survey I-D is ignoring) and that we're not chasing a
>"pie in the sky".
>
>>> If one or more such protocols exist, I don't see why they aren't
>>> discussed at this stage.
>>
>> I do not think that it is in our charter to analyze *all* possible
>> protocols.
>
>No, but if a good such protocol indeed existed, IP or not, having it
>before us might just help quell skepticism.
>_______________________________________________
>Roll mailing list
>Roll@ietf.org
>https://www.ietf.org/mailman/listinfo/roll
_______________________________________________
Roll mailing list
Roll@ietf.org
https://www.ietf.org/mailman/listinfo/roll