> > Better yet, make them PVM clients over the CTC, and map that PVM
session
> > as /dev/console. Then you get all the cool automation stuff and
> > multisession support of PVM as well.
> 
> What kind of terminal are you sitting at? Seems to me if you're
starting
> at a block-mode device you have problems.

Well, actually not. 

After doing some digging around and by grace of friends with archives of
ancient goodies (thanks, Arty!), the PVM link signon and session
establishment protocols are really the only part that implies any
structure on the communication. Once the link is active, it essentially
boils down to passing data without translation (similar to the way
tn3270 works -- negotiate a clean bitpipe, and then send raw 3270 frames
back and forth over the connection, interpreting them at the endpoints
as needed). PVM doesn't actually care what the data it's passing is,
once the session is up and running -- you wouldn't even have to
translate it to/from EBCDIC. 

The proposed console driver modifications would need to initialize the
CTC connecting to PVM, initialize the link and signon, and then wrap the
I/O to the console into PVM data packets. For completeness (and to keep
PVM happy), the modified driver should also shut down the link and sign
off properly at system shutdown. It might be a bit jerky due to the
packetization of the I/O within the PVM protocols, but it won't be any
worse than doing it over a console switch. 

With some slight adaptation of the way the 7171 provided a way to switch
into a "transparent" mode, I think one could get a console stream that
would permit what Brandon wants w/o re-engineering the world. What I
haven't figured out yet is whether this would require substantial
surgery to the client, or whether we'd be better off putting that
surgery into a proxy server and doing it all once. 

The PVM internode protocol is quite sophisticated -- the designers
planned very well for doing stuff like this. If we could get a Linux PVM
daemon to communicate with PVM, and then do session mapping in that
daemon, then we'd be home free without inventing lots of polyhedral
wheels. This would also be a convenient way to be able to move the
session mapping outside the 390 if we chose the PVM over IP driver as
the communication to the console. 

Another thought is perhaps to use the existing LAT client and daemon as
the starting point, since it already has the kind of session mapping
code we want, and just rip out all the DECnet code and replace it with
stuff to handle the PVM sessions. That could get us there quite quickly,
and latd already has the concept of mapping specific terminals to
specific applications, which (with a combination of some entries in
inetd.conf) could allow 'telnet frontend xxxx' to map to a specific
console session with SSL protection, authentication, etc handled in the
frontend. 

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to