search

dcdolson / draft-dolson-sfc-hierarchical

IETF draft for hierarchical service chaining
1 stars 3 forks source link

Can number of options be reduced? #8

Open dcdolson opened 7 years ago

dcdolson commented 7 years ago

Too many options makes standardization less workable. Should section 3.1 be pivoted to discuss requirements of SFs, allowing the IBNs to use any method that don't break the SFs?

dcdolson commented 7 years ago

Section 3.1.3 -- unique paths -- is particularly limiting, since it requires consistency of metadata in upper and lower paths, and is incapable of stashing information and popping it on sub-domain egress (e.g., TTL).

boucadair commented 7 years ago

I would remove: “3.1.3. Using Unique Paths per Upper-Level Path” and probably “3.1.5. Stateful/Metadata Hybrid”. The limitations of the 3.1.3 are obvious. 3.1.5 has the limitations of both the stateful and MD insertion methods.

s-homma commented 7 years ago

If possible, I'd like to leave section 3.1.3. The approach doesn't require nsh in lower level, and thus would be useful for environment where NSH aware and unaware SFs coexists. Also I assume that the variation of NSH (comvination of SPI, SI and TTL) in upper level is not so many in hierarchical SFC, and IBN only has to assign per NSH including TTL to unique lower level paths.

However, I understand the limitation, and we should remove it if it is inferior to other approaches obviously.