Based on the robustness group teleconferences, RadVision has submitted a delayed contribution named: RV-Robustness.zip available at: http://standard.pictel.com/ftp/avc-site/Incoming/RV-Robustness.zip
In addition to this submission, here is an incomplete list of open issues for robustness. Most of these have been discussed in teleconferences or on the mailing list. Please send any open issues to the mailing list and we can discuss them further in Geneva.
Thank you.
Open Issues
1) How does robustness interact with billing? There are concerns about keeping a call alive that can't be billed for properly.
2) Does an endpoint need to reregister with the gatekeeper after it crashes and comes back up? When does it perform this process? Suppose endpoint A and B are registered with the same gatekeeper. After B crashes and it comes back up should it initially try to recover a call that is in progress? Then, if no calls are in progress, should go ahead with gatekeeper discovery or perform gatekeeper discovery immediately? This scenario raises security issues....
3) How does security interact with robustness? For example, does an endpoint need to re-authenticate when it (or a gatekeeper) crashes and comes back up? (answer should be "yes".) How? There are concerns that the robustness mechanism will introduce security holes.
4) What identifier (CALL ID???) will be used when H.225 and/or H.245 connections are reopened? RadVision's submission suggests we use the CALL ID.
5) How will the H.225 and H.245 connections be reopened? (This might belong in the implementers guide)
6) There has been considerable debate on whether or not we allow the H.225 connection to be closed during a call. Should we change the standard to say that it can't be closed?
7) The standard states that if the H.245 connection is closed then the call is terminated. How does this fit in with robustness procedures? Should we change this requirement in the standard?
8) (from the mailing list) If an endpoint fails to find a gatekeeper during gatekeeper discovery, it is permitted to assume that there is no gatekeeper around, and that it may therefore make calls without using a gatekeeper. It might be considered "polite", however, for an endpoint carrying on this way to periodically check up to see if there ARE any gatekeepers around. At least one well-known H.323 stack acts in this way if asked to handle RAS for its application. The question is what should it do if it should find a gatekeeper at this later stage?
9) Is support for robustness mechanisms in H.323v4 mandatory or optional?
-- maureen
Maureen Stillman Member of Scientific Staff Voice: (607)273-0724 x62 Nokia IP Telephony Fax: (607)275-3610 127 W. State Street Mobile: (607)227-2933 Ithaca, NY 14850 e-mail: maureen.stillman@nokia.com www.nokia.com