Thanks for your responses - see below

> On Jun 13, 2024, at 2:34 PM, Dino Farinacci <[email protected]> wrote:
> 
> 
>>>>> 
>>>>> What is the precedence between the L bit and the S bit? Must they both be 
>>>>> set to 1? What is the procedure of L
>>>> 
>>>> There isn't one, they are treated independently.
>>> 
>>> PPE
>>> How is the decision to skip a node taken?
>>> Let’s take this example  ( n (L=1, S=0, P=0), n+1 (L=1,S=1),…., etr)
>>> 
>>> If we check S=0 first then the decision is just to Skip that hop “n” then 
>>> will not do a look up and go to next
> 
> Right, if the encapsulator thinks the RLOC for n is down, it could go to n+1.

Here is where i wonder whether strict would have been best to drop the packet 
and not go to n+1 per the example for SFC where there are mandatory services.

I think it might be worthwhile to document this behavior so as there are no 
surprises.
Thoughts?

Padma
> 
>>> reencap hop n+1 . Get a valid RLOC by performing look up for next hops in 
>>> the list. (Result “n”is skipped).
> 
> Then it looks up EID n+1 to get the RLOC for it.
> 
>>> If we check first L = 1 and then there is a valid RLOC we use the reencap 
>>> hop n regardless of whether it was skippable
>>> (Result here “n” is not skipped)
> 
> Right.
> 
>>> If S =1 then we may do a look up (Result “n” not skipped)
>>> See more below on S=1 and unreachability.
>>> 
>>> I was wondering which is the actual behavior. As the text on S=0 is a “can 
>>> skip” it seems implementation specific and still be ok.  Should some 
>>> guidance on use of L and S bit be included for clarity?  FWIW I have no 
>>> strong opinion on this.
> 
> You have to lookup the EID when L=1, you can't tell the EID is up or not, you 
> have to probe its RLOC. When you do that, you can decide not to use it and 
> move to the second one. If the S=1 is set for this EID, then you must stop 
> using the ELP.
> 
> Dino
> 

_______________________________________________
lisp mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to