Notes from Friday, April 9 1999
Thanks to all those who participated the the very successful (and fun) BOOM demonstrations on March 30. The ITX demos actually stayed up and ran!
What still needs to be done: Buffering stuff in data exchange, RTP packet tuning, keep alive signals for detecting application crashes (Note: done as of 4/14), conference calls, and getting DTMF over the telephone, gateway has a thread that blocks and has to be fixed, and there are persistent thread crashes within the JTapi library. Another to-do is to make the gateway PBX-independent.
Upcoming video taping of the project (Friday April 16).
Yang has been working on text-to-speech using Microsoft's SAPI. Speech-to-text is also available in that package.
Jason has been looking into Kerberos for possible use with Abhideep's alarmclock project. (This application will let you set a time at which you want your telephone to ring to wake you up.)
Keshav said that data exchange should have about 10 msec latency when on a subnetwork (as in the system lab)! But we are at 125-200 msec (a big improvement over where we had been, by the way). Keshav mentioned "software slicing" as a technique for measuring latency and seeing where it comes from.
One win from our binding signaling with ITX applications is that the signaling component can take care of calling isActive on its peer and notifying the application appropriately if the peer has crashed. For example, you could invoke a "Dial" on your DesktopSignaling, and it will utimately return with a reject if the dialled peer has gone away for some reason. Likewise, if your application has accepted an incoming call, your onAbort method can be invoked if your DesktopSignaling detects death of a peer.
The data model is working very well. Donna was able to write a little phone recorder application in about 90 minutes and as many lines of code. Our API isn't bad!
Donna will order a Dialogic card with fax functionality on it, for James to use this summer.