At the interop, I sent an ARQ to a GK with an unregistered alias on purpose.
The GK scanned its database, couldn't find it, and then queried another GK. I
tried this maybe three times, and found that it took around 20 seconds for
"my" GK to respond with its ARJ. The recommended time out and retry count is 3
seconds and 2 times, respectively, for an ARQ (7.19/H.225.0v2). This adds up
to a maximum recommended time out of 9 seconds. The person testing the GK
thought that it should be longer due to the fact that a GK may have to query
another GK before responding. The GK-GK query should always be one-deep,
regardless of whether it uni- or multicast the LRQ, so the time-to-reply is
not indeterminate.
Should we increase the recommended total wait for an ARQ to around 30 seconds,
such as a time out and retry of 10/2 or 5/5?
I know that a GK can use RIP with a v2 EP to increase the timeout, but
increasing the recommended values would work with all EPs. Any thoughts?
Paul Long
Smith Micro Software, Inc.
