This has been the subject of some discussion between Mr. Lee's company
and ours over the past nine weeks (we are the affected clients in
question; as outsourcers to the personal lines insurance industry, we in
turn are supporting our own clients on this application), and has of
course resulted in my posting several questions to this list. We
involved HPRC in this issue (who involved the expert center), as well as
our data center's network people, and our client's network people. All
three involved parties simultaneously took traces of multiple instances
of this application, capturing working exchanges and failing exchanges.
HPRC also had our data center put a memory dump to tape, immediately
after some workstations failed. After almost a week of priority one
analysis of their TCP and link traces and the memory dump and a large
number of phone calls between our company and HPRC, the expert center
concluded that telnet and TCP are responding correctly to everything
they sees. We (my company) are now pursuing Time and Materials work with
HP (the earliest they can commit resources to this is in three weeks) or
a channel partner, and with those who replied to my "Seeking TCP
consultancy" posting, to troubleshoot this application. I have also
discussed this off list with more than one of the subscribers, who know
MPE telnet and tell me that it works just right, and have discussed
their availability to consult with us on this. I have also briefly
corresponded with the author of a book on WinSock, who developed the
WinSock DLL implementation for HP's TCP/IP stack, and participated in
the development of Microsoft's WinSock specification. We have double
checked our TCP settings in NMMGR, sharing them among all parties, and
tried tuning them to no avail, and are currently pursuing one more set
of concurrent traces.

It is worth noting that while our development and production platforms
are on MPE 55 Express 4 with PTDEDU6A, located across a WAN, the
application was (as I understand it) developed to go to MPE 50 on a LAN;
I do not know it's express level or patches applied. One party suggested
we double-check telnet option negotiation. Also, neither we nor our
vendor see this application fail across our WANs to the same 3K, and the
West coast client only sees the application fail during East coast
business hours.

If those on this list and in the know can offer means whereby either
company or both can troubleshoot this, we would both be immensely
grateful. I too welcome replies, off list to either or both of us if you
think MPE telnet works (I am satisfied thus far that it does, but of
course each involved party is confident that their environment is not
the problem), or on list if you regard this as an issue with the
integrity of MPE telnet or the inetd, and of likely interest to the
list.