Ada Validation

Ada Validation

[Kenneth Mays]

Subject: The Validation of Ada95

Greetings,

The Ada Joint Program Office (AJPO) runs the show in validating Ada95 
compilers. This issues
deal the Ada95 compiler (whether a true cross-compiler or basic 
compiler) being compliant
to the Ada95 RM. This means that the Ada95 compiler is not a subset 
or superset of the
Ada95 language - it is the standard version of the language.

Now, validation doesn't mean "bug-free". Validation only means it 
complies strictly to the rules
set forth by the AJPO's Ada Compiler Validation Capability (ACVC) 
model. For those of you who never read DoD Directive 3405.2, it 
states that only validated compilers may be used in mission critical 
systems.
Supposedly, you should see an emblem of a pentagon with the words 
"Validated Ada" within the pentagon. Now, who is to say that a vendor 
doesn't use the emblem to sell of a validated Ada83 compiler that can 
compile a subset or superset of Ada95 code?!? Since the DoD Directive 
was out in 1987, maybe we should change it to say "Validated Ada95" 
within the pentagon. Then, there is an issue of 100% compliant or  
98%. If you are going to use this spanking new Ada compiler for your 
mission critical system do you want a non-validated, bug-ridden, 
non-standard Ada something compiler in your embedded system. I think 
not.  You'd want your compiler to comply to some standard, and hope 
you didn't spend $500+ on some hack. Then again, validation doesn't 
mean bug-free - so understand that part as well. ALSO, if any 
compiler doesn't conform to the validation it must state in the 
documentation or program that it is a subset or superset of the 
language or just a hack). This protects (you hope) the user from a 
hardware dependent version (must use a certain library from a certain 
computer to work correctly (read JAVA)).

Basically, if you run code on your SGI platform with a validated 
Ada95 compiler - I can compile it ona Sun workstation with another 
Ada95 validated compiler. You shouldn't have to modify the program to 
et it to work - which makes it very portable across platforms (you 
would hope). If you use C++, you'd want to compile your nonhardware 
specific C++ programs using GCC V2.7.2 and have that same C++ program 
on AT&T C++ Release V3.x without a hickup. If you remember True 
BASIC/GWBASIC/BASICA/HP-BASIC then you might understand the 
difficulties of programming for different variants of a similar 
language.

For those of you who don't like standards (I'm not saying they are 
the answer to everything - but try SCSI on various platforms), then 
think of why ANSI/ISO, VESA, and other organizations got together. We 
can stick together an get GNAT-95 bug-ridden and Ada95 complaint for 
validation - since it is free and avalable on all systems. Then, we 
will have somewhere to start without breaking the bank.
-Ken

Re: Ada Validation

[Carl Bowman]

In article <00001a73+00002c38@msn.com>, Kenneth Mays <KMays@msn.com> wrote:
>Subject: The Validation of Ada95
>
>Greetings,
>
>The Ada Joint Program Office (AJPO) runs the show in validating Ada95 
>compilers. This issues
>deal the Ada95 compiler (whether a true cross-compiler or basic 
>compiler) being compliant
>to the Ada95 RM. This means that the Ada95 compiler is not a subset 
>or superset of the
>Ada95 language - it is the standard version of the language.
>
...
>
>model. For those of you who never read DoD Directive 3405.2, it 
>states that only validated compilers may be used in mission critical 
>systems.
>Supposedly, you should see an emblem of a pentagon with the words 
>"Validated Ada" within the pentagon. Now, who is to say that a vendor 
>doesn't use the emblem to sell of a validated Ada83 compiler that can 
>compile a subset or superset of Ada95 code?!? Since the DoD Directive 
>was out in 1987, maybe we should change it to say "Validated Ada95" 
>within the pentagon. 
>
...
>
...
>
>-Ken


Thanks for your observations, but there is a separate, very different
certification mark for Ada95 compilers.  The graphic is not a pentagon
and contains the "Ada globe" and the word "Ada".  Ada83 and Ada95 are
words we use to differentiate versions of the language - the name is
still Ada.  Both certification marks clearly state the standard by
which the compiler has been certified.


If I've misunderstood your post, my apologies.


Carl Bowman
Ada Information Clearinghouse

Re: Ada Validation

[Tore Joergensen]

Kenneth Mays (KMays@msn.com) wrote:
: For those of you who don't like standards (I'm not saying they are 
: the answer to everything - but try SCSI on various platforms), then 
: think of why ANSI/ISO, VESA, and other organizations got together. We 
: can stick together an get GNAT-95 bug-ridden and Ada95 complaint for 
: validation - since it is free and avalable on all systems. Then, we 
: will have somewhere to start without breaking the bank.
: -Ken

GNAT 3.00 for SGI is validated (I'm not sure how it works, but I guess
3.04 is validated too (?)). The other ports use more or less the same
code. If I've understood resent postings correctly, the reason for not
validating all the ports is that a validation is expensive (I think 
somebody mentioned twenty grands, which is much to spend on a free 
compiler). I've not had any problems with GNAT 3.01 for OS/2, and I think
it is a very nice place to start without breaking the bank :-). Free is a
very attractive price for students and hobby-programmers. BTW, it is
not available on all systems, just a fairly big bunch.
-- 
+-------------------------+-------------------------------------------+
| Tore B. Joergensen      | e-mail : tore@lis.pitt.edu                |
| Centre Court Villa      | web    : http://www.pitt.edu/~tojst1      |
| 5535 Centre Avenue # 6  |                                           |
| Pgh, PA 15232, USA      | Norwegian MSIS-student at Univ. of Pgh.   |
+-------------------------+-------------------------------------------+

Re: Ada Validation

[Robert Dewar]

Tore said

"Tore said

"GNAT 3.00 for SGI is validated (I'm not sure how it works, but I guess
3.04 is validated too (?)). The other ports use more or less the same
code. If I've understood resent postings correctly, the reason for not
validating all the ports is that a validation is expensive (I think
somebody mentioned twenty grands, which is much to spend on a free
compiler"


Tore, you have not understood these "recent postings", indeed I have
no idea what you are talking about. GNAT may be free, but of course
serious use of GNAT with support is not free, and users who rquire
validation are typically not casual hobbyists (who tpically do not
care about validation). 

Ada Core Technologies is planning on validating on all major targets.
We decided not to do any further validations under 2.0, and we are
now working on 2.0.1 validation.

Out validation schedule is not determined, and will be decided on the
basis of commercial considerations and requirements.

Note that the 3.04 that you pull off by anonymous FTP is not technically
validated, althugh it may be from the same code base as the validated
compiler. To have a compiler that meets formal validation requirements,
you must have a contract with one of the certificate holders, i.e.
from SGI or from Ada Core Technlogies.

Robert Dewar
Ada Core Technologies

Re: Ada Validation

[Tore Joergensen]

Robert Dewar (dewar@cs.nyu.edu) wrote:
: Tore, you have not understood these "recent postings", indeed I have
: no idea what you are talking about. GNAT may be free, but of course
: serious use of GNAT with support is not free, and users who rquire
: validation are typically not casual hobbyists (who tpically do not
: care about validation). 

I agree :-)

BTW, when I tried to find the posting I thought I had seen (about 20K),
I wasn't able to find it, so I was probably mixing together something...
I'm sorry about that (Well, I'm not sorry that it doesn't cost $20K :-).

: Ada Core Technologies is planning on validating on all major targets.
: We decided not to do any further validations under 2.0, and we are
: now working on 2.0.1 validation.

: Out validation schedule is not determined, and will be decided on the
: basis of commercial considerations and requirements.

: Note that the 3.04 that you pull off by anonymous FTP is not technically
: validated, althugh it may be from the same code base as the validated
: compiler. To have a compiler that meets formal validation requirements,
: you must have a contract with one of the certificate holders, i.e.
: from SGI or from Ada Core Technlogies.

At least I got some information I didn't know :-)
-- 
+-------------------------+-------------------------------------------+
| Tore B. Joergensen      | e-mail : tore@lis.pitt.edu                |
| Centre Court Villa      | web    : http://www.pitt.edu/~tojst1      |
| 5535 Centre Avenue # 6  |                                           |
| Pgh, PA 15232, USA      | Norwegian MSIS-student at Univ. of Pgh.   |
+-------------------------+-------------------------------------------+

Ada Validation

[Gregory S. Lakis]

Hello Everyone,
I have a question about Ada compiler validation.  To be specific 680X0
cross compiler evaluation.  We are currently using the Verdix 680X0
cross compiler for a variety of projects.  When we download the program
to our target environment which is a 68020 based "home grown" CPU board
is that code still considered to be "validated" Ada?  I have had a
variety of answers ranging from a straight yes, to "no, you have to
send your CPU board to the vendor for revalidation" which is totally
unacceptable for classified hardware.  Any opinions would be greatly
appreciated.

Greg Lakis
Raytheon - Submarine Signal Division
Portsmouth, RI 02871

gsl@rayssde.ray.com