Re: Troubleshooting interrupted Telnet connections

["James H. H. Lampert" <jamesl@xxxxxxxxxxxxxxxxx>]

Scott Klement wrote:

> The Cisco VPN client may be smart enough to ignore keepalives.

I'd say that in that case, the word "smart" ought to be in quotation 
marks. Smart as in equus asinius.

One thing I turned up was the possibility of an in-band alternative 
keepalive, involving periodic sending of IAC NOPs. (But can a TN5250 
server in a keyboard lock state handle IAC NOPs without being driven 
insane?)

In the proprietary protocols used by our products, instead of depending 
on keepalive probes (which I'd never heard of at the time, and which 
wouldn't have been as effective in any event), we solved the "orphaned 
child-server" problem with what I call a "heartbeat": the protocol is 
required to have a NOP request defined, and the client is required, when 
idle, to send a NOP every 30 seconds. If the child-server is waiting for 
requests, and doesn't get at least a heartbeat within 2 minutes of 
entering the "wait for requests" state, it assumes that the client to 
which it is connected has abended, locked-up, or gone insane, and it 
quietly "starves to death."

--
JHHL