[I've change the subject from "TaskManager progress" to reflect a change
in focus from TaskManager refactoring to understanding the SDM
concurrency behavior.]
You say "I had a hard time". Does it now work, in the sense of achieving
a reasonable balance between the multi-cast and uni-cast protocols? If
so, did you fix something in SDM or change how you were using it to get
it to work?
Patricia
On 8/6/2010 7:47 AM, Gregg Wonderly wrote:
Okay, it looks like there are several things that we need to look at
with the SDM code then(*). I think we need to look carefully at the
mutations of the data that are occurring and that River-324 tries to
help manage to see if we can get our heads around it and hopefully make
sure that we believe that River-324's problem of stale or wrong cache
contents can be managed more readily.
Gregg Wonderly
* I had a hard time getting SDM to work reliably for me, and in
particular, it initially did not perform unicast lookups, but waited for
multi-cast announcements only and this made it non-functional on
non-local network segments.
