Notes from eVLA software meeting of April 10.

B. Sahr called attention to the recent E-Mails from W. Koski and B. 
Carlson about document numbering.  Ranges and types are given for 
software documents.  Exact assignment of numbers probably require
coordination with BS or BC, respectively.

B. Clark made remarks about the nature of the Jbed RTOS.  It seems to 
be oriented to single user using a complex device.  Things that arise
in that context are facilitated.  Our model is slightly different, and
some things we need are clumsey.  A java implementation of a semaphore
seems to run to about a page of code, and may only be callable from tasks,
as opposed to threads.  (A semaphore or equivalent is needed to get 
control of the SPI bus.)

There was a brief discussion of security, to little end.  Clearly,
the system must offer the capability of 1) within the site, seeing 
MIB values and pushing buttons without much interference from a security
mechanism, and 2) beyond the firewall, being able to see MIB values
and push buttons after some sort of password security.

B. Sahr reports on his continuing investigation of ThreadX and Nucleus+
ooperating systems.  A basic operating system, with an IP stack, can
be put together in anywhere from 150 to 270 kBytes.  Stacks are available
both from the RTOS vendor and from two outside vendors.  Various other 
add-ons are available - a file system (do we need this?); http server,
JVM, etc.  Rough costs are about $15 K for the OS, an additional $15 K
for the IP stack.  HTTP server costs from $10-25 K depending on features.  
Tool suites are available from Green Hills (supports both, but better 
integrated with ThreadX) and Altium Tasking (supports both, but better
integrated with Nucleus+).  Altium has specific support for the TC-11 IB.
Cost of the tool sets will run about $7 K per seat.

Question was raised whether we continue to look at Jbed.  B. Clark feels
that it would take significant effort to conclude whether that approach
will do what we want it to do, and that we should commit the necessary
couple of man-months only if we feel that the system has significant 
advantages over the approach of a JVM layered on a more conventional 
OS.  B. Sahr feels that it is a major disadvantage that it is likely to 
be a more difficult system for the hardware people to work with.  K. Ryan
will try executing a C function from java, just to see if it is as easy
and convenient as he thinks it might be.