TamTam roundup.

Hal Murray hmurray at megapathdsl.net
Fri May 18 22:14:30 EDT 2007

> Could we use ntp on each system to keep the whole mesh in sync time
> wise, and just pass time stamps? 

How closely do you need the times to track?

Can you easily patch the current software to add a deliberate offset and see 
how it sounds?  Try things like: 10 ms, 30ms, 100 ms

The speed of sound in air is 1 ft per ms so a 10 ms offset will be the same 
as moving the laptop 10 feet (3 meters) closer or farther away from the 

> Assuming that things are synced closedly enough, that should make it
> work even with enough load on the network to up the latency. 

I've been assuming that the school server would be a NTP server and that the 
XOs would sync to it.  Unless the network is loaded or flakier than I expect, 
the XOs should all track the school server "pretty well".

I think we'll have to do some experiments to find out if it's good enough.  
Ping data would be a good start.  Can you get some on a typical (ha) network, 
both when idle and when running TamTam and some other network intensive 
programs?  I'd expect it to work if ping times are usually under 20 ms.

NTP assumes that the network delays from client to server and back are 
symmetric.  It gets confused if that's not true, but that won't bother TamTam 
as long as all the XOs see the same asymmetry.  I'd expect troubles if the 
network delays vary a lot, say from 10 ms to 100 ms depending on the load.  
If that only happens occasionally, we can probably configure ntpd to filter 
it out.

These are my opinions, not necessarily my employer's.  I hate spam.

More information about the Devel mailing list