SNTP - correct loop problem

I'm creating this RFE in response to the customer problem reported in the PMR. The customer's SNTP server ended up in a loop with another z/OS SNTP server on another system.

It appears our SNTP can get into a loop sending responses back and forth between two or our SNTP servers when our server receives a packet from another server instead of from a client. I looked at the following SNTP RFCs (each RFC pertains to a specific version of NTP, Version 2 through Version 4):
NTP V2 - RFC1119
NTP V3 - RFC1305
NTP V4 - RFC5095

We say our SNTP is V4 but we don't support associations and we don't support symmetric mode. As far as I can tell, our SNTP function only has the following support:
- Client/Server: receives requests and responds with a timestamp as a server
- Sends out multicast/broadcast packets with a mode of broadcast and a timestamp.

The customer figured out who sent the bad packets and stopped that server so they no longer have the problem. They are not interested in testing a fix.

So we are creating this RFE to fix the problem in a future release. The fix should change the Client/Server support code to only process input requests where:
- the mode is Client or,
- there is no mode value (supposedly this is possible with pre-V2 versions of NTP) and the source port is not 123.

All other input requests would be silently discarded (but we could trace them). There is already code that skips input requests with a mode of broadcast, so maybe we could just update this code. We should also change the IP Config Guide documentation for SNTP to indicate that it does not support symmetric mode.