Re: [mpls] PSD only -- WAS:(Re: Two week poll on the use/no use on an unique MNA PSD label)

[Loa Andersson <[e-mail]>]

Oooopppsssss!

Sorry I found the "Post Stack Network Action Offset" discussed in 
draft-jags-mpls-ps-mna-hdr, and yes that is where it should be.

But it goes into a format B LSE, using OpCode 1 + offset counted in four 
octets.

As I see it this gives the same 64 bit overhead as before, and LSE A, B 
and C are for all practical purposes "ISD" even if they are discussed in 
the "PSD" draft.

This actually beg the question what you do by the offset if it is not 
coded in the ISD?

/Loa

On 2023-05-21 12:07, Loa Andersson wrote:
> Tony,
> 
> I have two questions inline.
> 
> On 2023-05-15 18:08, Tony Li wrote:
>>
>> Hi Loa,
>>
>>
>>>>> Is this an alternative method to indicate PSD, without any other ISD?
>>>> I agree with their answer.  If you want to indicate that there is 
>>>> PSD using ISD, there is a way to do that.
>>>
>>> OK, that is fine. As I understand it this would be 32-bit overhead? I 
>>> don't think that is too problematic to be the preferred method.
>>
>>
>> I think that you would need a Format A LSE for the bSPL and a Format B 
>> LSE with the Post Stack Network Action Offset.
> 
> I'm looking in draft-ietf-mpls-mna-hdr, and can't find "Post Stack 
> Network Action Offset", where is it defined?
> 
> Also if this is "PSD only", why would there be an offset?
> 
> /Loa
> 
> 
>>
>> That’s 64 bits of overhead, plus the additional processing at each 
>> transit node to parse and evaluate the ISD and the PSD and then to 
>> realize that it has no work to do.
>>
>>
>>> "if the control plane indicated the presence of PSD", are you 
>>> proposing a new PSD FEC?
>>
>>
>> No. We’ve made the point the control plane is supposed to distinguish 
>> our payload and thus it’s ok to overload the first nibble of PSD.  If 
>> that is the case, then the control plane should be able to indicate 
>> that we carry MNA PSD (as well as other PSD), plus the address family 
>> of the payload.  If we have that indication, then there should be no 
>> need for an ISD indication of PSD.
>>
>>
>>> I said before, if you have the clear context in the label stack, you 
>>> don't really need the first nibble.
>>
>>
>> We lack clear context in the label stack for other types of PSD. Why 
>> incur unnecessary overhead? Why not be consistent with what’s already 
>> there?
>>
>> Tony
>>
> 

-- 
Loa Andersson                        email: [e-mail]
Senior MPLS Expert                          [e-mail]
Bronze Dragon Consulting             phone: [phone]