Re: Critique of SEI's Ada DARK project

[spray@convex.com (Rob Spray)]

In <1991May22.143229.24667@src.honeywell.com> vestal@SRC.Honeywell.COM (Steve Vestal) writes:


>In article <spray.674858011@convex.convex.com> spray@convex.com (Rob Spray) writes:

>Rob> One concern I have always had about DARK is how it got funded
>Rob> targetted to a processor that is not available in a 
>Rob> radiation-hardened configuration.  Several embedded systems
>Rob> I am familiar with are required to use full MIL-SPEC processors.

>Does mil-spec mean that a processor must be radiation hardened, or just that
>it be class B, certain temperature range, certain packaging, etc.?  "Radiation
>hardened" isn't a yes/no thing; there are several parameters and several
>degrees to radiation hardening.  For example, of the two processors cited, I

There's a MIL STD (5400 Class 1? I can't remember) that specifies
hardness, altitude, temperature range etc, etc.  It's my understanding
that most air/spaceborne MCCR systems must use processors that meet that
Standard (e.g. Loral(Rolm) Hawks, AN/UYK-xx, ap101, 1750A etc)

>orbit, shuttle or space station in low orbit, and aircraft, all vary.  Also,
>this is a systems issue, since one can trade shielding against the radiation
>insensitivity of the circuitry used.

Not always true. Most RFPs specify the Standards to be met.
Shielding non-qualified parts is not always an option.

>There are Ada development environments and executives for rad-hard computer
>systems, but I don't think a high degree of radiation insensitivity is a
>universal requirement on all military systems.

Agreed, but a significant number of systems have that requirement, 
so the choice of the 68K for DARK, effectively ignored many issues
faced by builders of such systems. 

The 68K has a reasonably orthogonal architecture and good memory
management scheme for distributed processing.  From a compiler
writer's view, lessons learned on a VAX transfer to the 68K.  So
the 68K had the best Ada environment to implement DARK.  (I 
suspect that proposing a VAX target would have been even more
unacceptable.)  However, there are no rad-hard 68Ks, so technology
transfer of DARK to a significant percentage of the intended systems
is hindered.

--Rob Spray
--spray@convex.com
--Previous disclaimer is still in full effect.