39 lines
1.3 KiB
Plaintext
39 lines
1.3 KiB
Plaintext
|
|
Potential upgrades to Channel Access
|
|
|
|
o generalized access to abstract data
|
|
o use multicast for broadcasting and ax the repeater
|
|
o improve client API
|
|
o improve security
|
|
o name service (wildcard queries)
|
|
o PV name prefix (for all PVs)
|
|
o client access to process passive and for maximize severity
|
|
o detect name conflicts at boot time
|
|
o multi priority connections (quality of service)
|
|
o reduce protocol overhead
|
|
o compressed protocol
|
|
|
|
|
|
o If there is a beacon anomaly then this indicates that
|
|
the server is _not_ available. Perhaps we should only
|
|
reset the search timer interval when we see a beacon
|
|
anomaly transition into a sure steady beacon.
|
|
|
|
o make certain that a monitor callback canceling itself
|
|
does not deadlock.
|
|
|
|
o the free list library does not now cause exceptions to
|
|
occur (it uses new ( nothrow )), and therefore we should
|
|
change the new handlers to be nothrow also if this is
|
|
the design goal.
|
|
|
|
o test the library when an IOC is running low on memory.
|
|
|
|
o The new CA interface should be multi-threaded by default.
|
|
We should continue to support a single-threaded interface,
|
|
but this would should be restricted so that it always
|
|
runs with preemptive callback disable and no threads
|
|
may join in. Alternatively, the new CA interface could
|
|
be 100% preemptive multi-threaded and the old interface
|
|
could be layered on this.
|