On Thu, Oct 24, 2013 at 3:56 PM, Reinaldo Penno (repenno) <[email protected]
> wrote:

>
>
>   From: "Y. Yang" <[email protected]>
> Date: Thursday, October 24, 2013 12:51 PM
> To: Cisco Employee <[email protected]>
> Cc: Wendy Roome <[email protected]>, IETF ALTO <[email protected]>
>
> Subject: Re: [alto] Problem with "longest prefix" rule for mapping
> endpoints to PIDs
>
> On Thu, Oct 24, 2013 at 1:23 PM, Reinaldo Penno (repenno) <
> [email protected]> wrote:
>
>>  I disagree again.  If you have a prefix 0.0.0.0/0 in one PID does that
>> mean no other prefix can appear in any other PID? it means that all other
>> prefixes will overlap with it.
>>
>
>  Not sure I understand the issue. Which condition among the three:
> (1) No prefix should belong to two PIDs.;
>
>  [RP] Wendy pointed out that this is already in the draft. Thanks for
> that. But I'm not sure this is necessary, but that is for another time.
>
>
Good statement. No-duplicate is not a necessary condition, but only a
sufficient condition. The wording in the current draft is "can", as
Sebastian pointed out. We often use "can" to imply "only sufficient" in
such a setting. It is not necessary when one considers the following case
(the duplicated prefix is dominated by a more specific prefix both PIDs).
But for now, I am fine with Sebastian's wording of enforcing the sufficient
condition.


>     (2) One obtains the union of all prefixes defined by all PIDs; call
> this set of prefixes P; (3) The longest prefix matching an IP address is
> identified in P. (4) The PID containing the prefixed identified in (3) is
> the PID of the IP address.
>
>  [RP] What is needed to be said? Above is already in the draft. The way I
> see it the draft already has everything needed.
>
>
> The current draft already has everything. Putting the conditions together
is to make potential issues more explicit, and can make the draft easier to
read.

Richard


>  Is it (1)?
>
>  Richard
>
>
>>
>>  The prefixes below could have been imported from BGP at t=0 and at t=1
>> they will be broken up. Or the PIDs can have different properties (as we
>> discussed before).
>>
>>
>>   From: "Y. Yang" <[email protected]>
>> Date: Thursday, October 24, 2013 8:47 AM
>>
>> To: Wendy Roome <[email protected]>
>> Cc: IETF ALTO <[email protected]>
>>  Subject: Re: [alto] Problem with "longest prefix" rule for mapping
>> endpoints to PIDs
>>
>>   Wendy,
>>
>>
>> On Thu, Oct 24, 2013 at 10:33 AM, Wendy Roome <[email protected]
>> > wrote:
>>
>>>  I'm glad to see I can still raise up a controversy!
>>>
>>>
>>  It is a very good discussion!
>>
>>
>>>  Here's the problem I saw. Consider two network maps:
>>>
>>>  MapA:  A1 = 10.0.0.0/14
>>>                A2 = 10.0.0.0/15
>>>
>>>  MapB:  B1 = 10.0.0.0/16, 10.1.0.0/16, 10.2.0.0/16, 10.3.0.0/16
>>>                B2 = 10.0.0.0/15
>>>
>>>  PIDs A1 and B1 cover exactly the same set of endpoint addresses. So
>>> the naïve view is that those two network maps are dentical.
>>>
>>>
>>   But if we strictly apply the "longest prefix match" rule, they are NOT
>>> identical.
>>>
>>  In MapA, 10.0.0.0 is in A2, while in MapB, it's in B1. To me, that's
>>> counterintuitive.  If y'all like that, fine. I can live with that!  But it
>>> should be documented, because I think this has serious potential for
>>> misinterpretation.
>>>
>>
>>  Yes. We should document it clearly in the document. The semantics is
>> that:
>>
>> (1) No prefix should belong to two PIDs.; (2) One obtains the union of
>> all prefixes defined by all PIDs; call this set of prefixes P; (3) The
>> longest prefix matching an IP address is identified in P. (4) The PID
>> containing the prefixed identified in (3) is the PID of the IP address.
>>
>>
>>>
>>>  And the next bake-off should have a test case for this -- eg, the
>>> network map should define PIDs like B1 and B2, and there should be a test
>>> that 10.0.0.0 is in B1, not B2.
>>>
>>>
>>  Good idea to test this in the next bake-off!
>>
>>  Richard
>>
>>
>>>  - Wendy Roome
>>>
>>>   From: "Y. Richard Yang" <[email protected]>
>>> Date: Wed, October 23, 2013 18:47
>>> To: Sebastian Kiesel <[email protected]>
>>> Cc: IETF ALTO <[email protected]>, Wendy Roome <[email protected]>
>>>
>>> Subject: Re: [alto] Problem with "longest prefix" rule for mapping
>>> endpoints to PIDs
>>>
>>>  These are fast responses. It seems that the majority (3:1) is to keep
>>> the non-merging semantics. If we do not hear other objections, we will keep
>>> the current version, but do clarify using the good example from Wendy.
>>>
>>> Cheers,
>>>
>>> Richard
>>>
>>
>>
>
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to