🚀 go-pugleaf

There are essentially three classes of TCP/IP products for MVS (actually
IBM systems in general).  The first is to put actual TCP on the MVS
machine and use some sort of channel attached Ethernet interface.  The
second, which works mostly for terminal (TELNET) traffic, is to build
an Ethernet protocol converter that plugs emulates an Cluster Controller
to the IBM system.  The converter accepts telnet connections and generates
a emulates a 3270 for the user.  The third is a box that sits on the
Ethernet that does most of the protocol work but requires special hooks
on the MVS side to interface it.

Emulators:

MITEK:  This is probably the nicest protocol converter I've seen for an
  IBM system (Ethernet or otherwise).  It has a hot key that gives you
  a nicely formatted screen telling you where all the funny 3270 keys
  are bound.  We didn't get around to trying any of the esoteric FTP
  or printer modes.   The method to configure the box could be a little
  nicer though.  User can add new terminal types easily (they are infact
  termcap entries from UNIX).

ACC:  You list a product similar to the MITEK that I have little experience
  with.

BRIDGE:  We got one of the bridge products CS1/SNA relatively soon after
  they were announced.  We had some difficulty getting it to work (probably
  mostly due to the preliminary version of the manual that we were using).
  We mostly panned it because it didn't support the most popular terminal
  we have on campus and the user could not add additional terminal
  descriptions).  This was all based on a test we made of it about a year
  and a half ago.

SUN:  Sun makes a board and software to go with it that plugs into one
  of their workstations to do the emulation.  The major problem with the
  channel attached version we have is that it crashes the MVS system it
  was plugged into.  Sun offered to buy it back, but has so far not bothered
  to make good on the offer.

REAL TCP/IP:
  The folks at UCLA did an actual MVS TCP/IP port.  This product usually is
  referred to as "UCLA TCP" but is sold under the commercial names of
  ACCESS/MVS etc..  Just about every TCP implemnation for MVS with the
  exception of IBM's is based on this.

  IBM TCP/IP for MVS.  This is a port of the VM product ("FAL") to MVS.
  It was done by creating enough special glue to wedge the VM stuff to
  MVS.  They even wrote VMCF for MVS to do this project.  It looks fairly
  good.

REAL ETHERNET Interfaces:
  ACC:  This was what the UCLA code was originally written for.  The ACC
  9310 is a box that plugs into an Ethernet on one side but emulates an
  IMP to the IBM host.  We had serial number 1 for a long time.  I think
  we have got number 17 now.  Ever since we got rid of box #1 it has been
  reliable.

  BTI:  (Bustek) makes perhaps the best performance box for an IBM host.
  It is regarded highly by the community.

  IBM 8232:  This box is an overgrown IBM PC/AT with a channel attachment
  card and an Ungermann/Trout PC/NIC Ethernet card.  It's major problem
  is that it is slow.  It's even worse with IBM's token ring interface.
  It is even slower than the performance of normal PC's running DOS
  doing TCP with the same hardware.  The thing comes rack mounted in
  a white cabinet mostly so that it looks mainframish and so that the
  thing doesn't fall over when some CE comes yanking on the channel cables
  to see where they go.

PROTOOCOL BOXES:
  Fibronics:  AKA Spartacus.  This box does most of the TCP/IP protocol
  work and connects to sofware running on the host.  I have little experience
  with this, but I would expect that in this day and age it is going to
  be a little obsolete.  The problems are likely going to be old versions
  of box firmware not supporting the protocols properly.

RECOMMENDATION:  Buy a BTI box and wait for the IBM TCP/IP software
  to come out, if you want full functionality.  If you only want terminals,
  buy the MITEK.

-Ron