ada certification question

ada certification question

[Brian L. Kahn]

:

I have a question about DoD certification of ADA.

I was recently told that certification is given for an ADA compiler,
it's runtime support, running on a certain machine with a stated
configuration, and targeted for a certain machine with a certain machine
with a stated configuration.
Thus a certified ADA cross compiler running on a Vax under UNIX and
producing code for a SUN workstation would have to be recertified if 
the code was to be used on some other 68000 machine.

How much of this is true?  Sounds outrageous.

If this is largely true, doesn't it mean that ADA can never be used on
a new machine?  A target machine must be at least as old as the certification
process is long?

Thanx, B<

Re: ada certification question

[Stephen Tihor]

blk@syteka.UUCP (Brian L. Kahn) writes: 
	... certification is given for an ADA compiler, it's runtime support, 
	running on a certain machine with a stated configuration, and 
	targeted for a certain machine with a certain machine with a stated 
	configuration.

	Thus a certified ADA cross compiler running on a Vax under UNIX and
	producing code for a SUN workstation would have to be recertified if 
	the code was to be used on some other 68000 machine.

	How much of this is true?  Sounds outrageous.

All of it is true.  Doesn't sound outrageous to me since any change in the 
target machine enviornment will require changes in the output of the compiler
or the run-time library or kernel in all but the most pathological cases.
Given what Ada certification is all about certification of the exact 
configuration of compiler host and target processors and O/S's is clearly 
desriable when producing important executables.

	If this is largely true, doesn't it mean that ADA can never be used on
	a new machine?  

Huh?
	A target machine must be at least as old as the certification process 
	is long?
True. Certification takes about a week plus paper shuffling under ideal 
circumstances. [Observed data from many many many runs of the ACVC tests 
here at NYU.]  Re-certification should approach ideal circumstances since
all the support tools and procedures that had to be worked out before
can just be used again.  New certification will also get faster as the 
APSE or at least some significant set of CAIS-implementation are available
and the tools need to certify are built in Ada.

Re: ada certification question

[jbn]

      The AJPO does not seem to be a bottleneck here.  In practice, compiler
vendors seem to pass certification for the first time long before they have
a usable product, since you can certify before you make your compiler either
fast or user-friendly.

					John Nagle

Re: ada certification question

[Norman Diamond]

>       The AJPO does not seem to be a bottleneck here.  In practice, compiler
> vendors seem to pass certification for the first time long before they have
> a usable product, since you can certify before you make your compiler either
> fast or user-friendly.
> 					John Nagle

Shouldn't this be posted to net.lang.fortran or net.lang.lisp?
After all, net.lang.ada is for fortran-vs-lisp debates.
-- 

   Norman Diamond

UUCP:  {decvax|utzoo|ihnp4|allegra}!watmath!watdaisy!ndiamond
CSNET: ndiamond%watdaisy@waterloo.csnet
ARPA:  ndiamond%watdaisy%waterloo.csnet@csnet-relay.arpa

"Opinions are those of the keyboard, and do not reflect on me or higher-ups."