El 26/06/2006, a las 13:13, Jari Arkko escribió:
On point b, I'm not sure there is a lot we can do. At least not without revisiting some of the requirement decisions, such as the need to support at least some scenarios where there's only unidirectional connectivity. Thoughts?
i guess that the main reason to support unidirectional connectivity is ingress filters. I mean, as far as i understand, failures affecting only one of the directions in a path are kind of rare, so it probably the added complexity is not worth it. However, if no additional mechanisms for dealing with ingress filtering are availalable, I guess that unidirectional connectivity can be quite common.
Now, the solutions to deal with ingress filtering seem to also imply quite a few changes in the sites, and it is not so straight forward that sites would like that.
So i guess that the options are either complexity to deal with ingress filtering or complexity to deal with unidirectional paths...
OTOH, ingress filtering compatibility is also required for non-shim capable hosts within multihomed sites...
So, i think we should have at least the option to support unidirectional connectivity, but maybe we could define also a simpler protocol that only verifies bidirectional connectivity for those sites that do implement ingress filtering compatibility mechanisms
Regards, marcelo
--Jari