Hello authors,
When implementing the C-SID mechanism defined in the draft, I find out
something is missing and something can be enhanced in the draft, so I provide
comments below.
1. The pseudo code does not align with original drafts
* NEXT-C-SID: in original draft [1], the node checks the DA.Argument.next
(the next C-SID in DA) instead of DA.argument in this draft.
* REPLACE-C-SID: in original draft[2], the if node checks the
DA.Argument.SI instead of DA.argument in this draft.
We have implemented and deployed following the original draft, so It will be
better to align with the code. We think use DA.Argument.next and DA.argument.SI
is correct, easy for implementation and hardware friendly.
1. The combined flavor pseudo code can be enhanced. The current version is
too complicated, and not friendly for hardware implementation. Actually, we can
simply combine the pseudo code of NEXT-C-SID and REPLACE-C-SID to make it very
simple. This is the same way to combine flavors such as PSP, USP and USD (to
generate PSP&USP&USD).
"Inserting the pseudo code of NEXT-C-SID flavor with DA.Argument
replaced by DA.Argument.Next between lines S01 and S02 of the REPLACE-C-SID
pseudo code. "
When DA.Argument.Next is not 0, the endpoint node performs NEXT-C-SID
processing to shift the Argument to the left by the length of the C-SID.
When the DA.Argument.Next is 0, the endpoint node performs REPLACE-C-SID
processing to replace the C-SID from Segment list to DA. Doing this can bring
a lot of enhancement to the whole draft, and bring benefits to implementation
and deployment.
1. For REPLACE-C-SID flavor, it will be good to check the next CSID is 0 or
not before updating the CSID to the DA when SI > 0, so that we can handle the
processing error when the REPLACE-C-SID flavor SID is the last C-SID in the
CSID sequence. For example, when the SI >0 and next CSID is 0, meaning the end
of the CSID sequence, so update the next 128-bit SID to the DA when SL >0, or
stop processing when SL=0.
These are some enhancements of the C-SID mechanism I found in implementation.
Hope we can update the draft to address the comments to help implementation and
inter-op test, thanks.
Best regards,
Qiangzhou
_______________________________________________
spring mailing list
spring@ietf.org
https://www.ietf.org/mailman/listinfo/spring
