On Mon, Oct 31, 2011 at 9:29 AM, Y. Richard Yang <[email protected]> wrote:
> On 10/31/11 11:27 AM, Richard Alimi wrote:
>>
>> On Mon, Oct 31, 2011 at 7:51 AM, Vijay K. Gurbani<[email protected]>
>>  wrote:
>>>
>>> On 10/31/2011 08:49 AM, Bill Roome wrote:
>>>>
>>>> I don't have complaints, but I do have an observation: A client
>>>> cannot distinguish between "pids are invalid" and "pids are valid but
>>>> this server doesn't have reliable cost data for that pair."
>>>>
>>>> Does that bother anyone?
>>>
>>> The problem, at least as I understand it, is not predicated on a PID
>>> being valid or invalid.
>>>
>>> The validity (or lack thereof) of the PIDs is a function on whether
>>> they appear in the network map and cost map.  Presumably, if they
>>> appear there, the PIDs are valid.
>>
>> I agree with Vijay and Ben that looking at the Network Map should be
>> sufficient (presumably the ALTO Client already has it in order to
>> interpret the Cost Map).
>>
>>> However, the greater question is this: does the ALTO provider know
>>> authoritatively the cost between all PIDs?  That is the issue
>>> under discussion.
>>>
>>> Clearly, the ALTO provider cannot know the costs between all PIDs.
>>> Again, consider our sample topology we used for the bakeoff [1].  In
>>> that topology, the ALTO provider had authoritative costs for
>>>
>>> mypid{1,2,3} ->  mypid{1,2,3} +
>>>                peeringpid{1,2} +
>>>                transitpid{1,2} +
>>>                defaultpid
>>>
>>> However, asking the ALTO server to provide a cost between defaultpid
>>> (Internet) and transitpid1 is useless, since at best, the server will
>>> be indulging in a guess (and at worst, it'd be lying).
>>
>> I agree - so, just to be sure I'm understanding correctly, do you
>> think that letting the Cost Map be a sparse matrix solves this issue?
>> Or do you think there is something else that should be done?
>
> In a "typical" understanding, a sparse matrix means that the missing
> elements have a value of "0" and hence do not need to be explicitly
> represented. In our context, then we may need to make clear the
> semantics that  the cost is unknown or not revealed. This can be
> important in that a client may use a sparse matrix library (database)
> which may use the "standard"  interpretation and then generate some
> wrong results.

Yes - good point. The actual language in the draft (to be posted
today) is along the lines of omitting entries that are not defined by
the ALTO server -- "sparse" is not mentioned at all.  I should have
been more specific in my reply above.

Rich

>
> Thanks.
>
> Richard
>
>
>> Thanks,
>> Rich
>>
>>> [1] http://www.ietf.org/mail-archive/web/alto/current/msg01147.html
>>>
>>> Thanks,
>>>
>>> - vijay
>>> --
>>> Vijay K. Gurbani, Bell Laboratories, Alcatel-Lucent
>>> 1960 Lucent Lane, Rm. 9C-533, Naperville, Illinois 60566 (USA)
>>> Email: vkg@{bell-labs.com,acm.org} / [email protected]
>>> Web:   http://ect.bell-labs.com/who/vkg/
>>> _______________________________________________
>>> alto mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/alto
>>>
>> _______________________________________________
>> alto mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/alto
>
> _______________________________________________
> alto mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/alto
>
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to