Pekka Nikander wrote:
On the other hand I'm not sure if you classification above is exact. Primary locators are used in SCTP, but do we need them in shim? Or did you mean "one current locator"?
I don't understand the difference.
Looked at RFC 2960 again... "primary" appears to be current selection in SCTP. Sorry, I mixed it with the application level defaults that are provided in SCTP. Yes, a single current path is a good idea.
I am not so sure about the usefulness of a generic bypass mechanism. Diagnostic applications will require such bypass, for sure, and it will probably be a useful mechanism during a transition period, e.g., for SCTP.
However, from an architectural point of view I think that the shim layer is exactly the right layer to share information about the available IP addresses. The set of addresses is a property of the host, and therefore seems to belong to the upper part of the IP layer. The more specific properties of a path (an address pair), like QoS and what protocols/ports are available, seem to belong to a upper layer, though. Hence, from this point of view IMHO the shim should take care of distributing information about the addresses that belong to the same host while, in general, path discovery functions belong to the upper layers.
This was reflected also in the other mail with the "stepped" protocol model, where there is an additional layer between the "basic shim" and the unmodified ULs.
"transition" (not sure how temporary its going to be...) and SCTP were what I had in mind. So I agree. But in the end that means that we need the bypass...
--Jari