El 16/02/2006, a las 20:26, Erik Nordmark escribió:
Geoff Huston wrote:In response to your query (and Erik can correct me if I'm wrong) It is my understanding in the specification that whenever the ULID pair is used as the locator pair then the SHIM6 extension header (the context value) is omitted from the packet. This allies to the initial contact pahse, and also in any subsequent time when the SHIM selects the ULID pair as the current locator pair.Correct. But folks have also suggested that we should allow the shim6 extension header all the time i.e., even if the receiver doesn't have to rewrite the locators. I don't see any harm in allowing this for folks that want their packets to be bigger...
FWIW, i think may be useful if we ever want to support prefix rewriting by exit routers. I mean, if we want to support this, the way i understand this could be, is that endpoints would establish a shim session between them and then they would let routers to rewrite source prefixes for those packets belonging to the established context. In that case, it would necesary for the ednpoint to identify those packets that belong to the shim context hence they should include the payload header even if they are using the ulids as locators.
So, i would suggest to allow this behaviour, or at least not explicitly forbidding it.
i.e. it is my understanding that the packet header is simply an implicit "please perform an address substitution" instruction to the other end, and the locator pair plus the context value identifies the particular substitution to a ULID pair that the remote end should perform. If the locators are the ULID then no substitution is necessary, and no extension header is necessary. At least thats the way I interpret the specification, and I was wondering if this should be explicitly stated in the spec.If we allow the above, it is really "please lookup the context state and check if you need to replace the locators", but since replacing the locator pair with the identical locator pair can be still viewed as a replacement, that's just a detail.Any ideas where such text would best belong in the specification?
imho it would be enough to make sure that the document does not precludes such behaviour...
OTOH, we could include a note in the Sending/receiving payload packets sections....
Regards, marcelo
Erik