Hi Jean-Louis,

Many thanks for agreement on so many points.

I have cut this down to just the open points. Please see in line.

Cheers,
Adrian

>> 4.2. Carrying the OF object in a PCEP message
>>  An OF object specifying an objective function that applies to an
>>  individual path computation request (non synchronized case) MUST
>>  follow the RP object for which it applies.
>>
>> I don't like the fact that the OF is sometimes here, sometimes
>> there.
>> You need to facilitate several different cases.
>>    a. Message contains just one request with an OF to apply
>>    b. Message contains several unsynchronized requests each with 
>>       an OF
>>    c. Message contains several synchronized requests with an OF
>>       to apply to the set of computations
>>    d. Message contains several synchronized requests with an OF
>>       to apply to the set of computations, and the message 
>>       contains one or more unsynchronized requests each with an
>>       OF to apply.
>> It seems to me that your handling of the synchronized requests is
>> a problem because it appears that you can have a separate OF for 
>> each request in the set, but have no way to say the OF that 
>> applies to the whole set.
>
> Not really.

:-)

>>    So, I think you need...
>>    <PCReq Message>::= <Common Header>
>>                       [ [<OF>] <SVEC-list>]
>>                       <request-list>
>
> No. Recall that an SVEC comprises a set of synchronized requests.
> A SVEC list is a list of SVEC, that is a list of sets of
> synchronized requests.
> We need to be able to specify an OF for each SVEC...
> So I think our proposed encoding is the right one.

Soooo.
I was assuming that you may need:
- an OF to apply to the set of synchornized requests
- an OF to apply to each of the requests

What you appear to have is:
- an OF to apply to each synchronized request
  through the OF in the SVEC-list
- an OF to apply to each of the requests
  through the OF in the request

It seems to me that you do not have the ability to supply a meta-OF that 
applies to the whole SVEC-list, but you have two separate ways to supply an OF 
for each request.

>>    where:
>>       <svec-list>::=<SVEC>
>>                     [<svec-list>]
>> 
>>       <request-list>::=<request>[<request-list>]
>> 
>>       <request>::= <RP>
>>                    <END-POINTS>
>>                    [<LSPA>]
>>                    [<BANDWIDTH>]
>>                    [<OF>]
>>                    [<metric-list>]
>>                    [<RRO>]
>>                    [<IRO>]
>>                    [<LOAD-BALANCING>]
>> 
>
> This is already what we have.

Well, I moved OF next to metric-list.

>> Now, you also have...
>>       <metric-list>::=<METRIC>[<metric-list>]
>> This is lifted from [PCEP] and is fine. But don't you also 
>> need a metric list for the synchronized OF?
>> This would yield...
>>    <PCReq Message>::= <Common Header>
>>                       [ [<OF>] [<metric-list>] <SVEC-list>]
>>                       <request-list>
>
> Right, this will be added

Unclear whether you are accepting my assumption that metric-list should be 
applied to the whole SVEC-list. If so, how is this different from the OF?

>> Now, suppose the request desires (not requires) an OF, and 
>> the response says No-Path. Shouldn't the response also say 
>> which OF was used to produce this result?
>
> Strictly speaking the reason for no path is never an
> objective function, the reason is a constraint.

Yes, that is very true.

But, nevertheless, isn't it helpful to report the OF that was used?

_______________________________________________
Pce mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/pce

Reply via email to