Print

Print


Allow me to repeat some questions I posted on
comp.lang.ada recently; these apparently got
a "blank stare".

> The Ada mandate (R.I.P.) was qualified thus:
> *validated* compilers shall be used.  So, there
> was an incentive for vendors to achieve validation.
> Now there is no mandate; I am assuming that this
> will result in project-specific decisions about
> the use of validated compilers vs. non-validated
> compilers.
>
> I ask the following questions to those who may know:
>
> Does the lack of a mandate mean that fewer projects
> will be requiring the use of validated compilers?
>
> If this is so, should we expect to see fewer
> compiler vendors seeking validation for their
> compilers?
>
> If so, is this because validation is currently
> so expensive to achieve that compiler vendors
> find that cost/benefit anlysis requires that
> they forego validation?
>
> If that is true, what can be done to make the
> validation process better/faster/cheaper; and/or
> what can be done to make validation more
> attractive for vendors and customers, so that
> a mandate is not required to motivate it?
>

A few motivations for these questions have
been in my mind for a while:

1) The Fall newsletter from AdaIC mentioned that
there are many more compilers than those on the
validated list.

> http://sw-eng.falls-church.va.us/AdaIC/news/Newsletter/1997/fall

2) Now that GNAT is freely available, a number
of organizations other than straight compiler
vendors might seek (actually I should say *have
sought*) to construct compilers for different
platforms than those supported by ACT.  If these
organizations (which may be small groups or even
individuals) want validation for the resulting
compiler for whatever reason, the "threshold"
may be too great to make it practical.

Some examples of home-grown compiler efforts
based on GNAT include work done at Honeywell
Defense Avionics (Matthew Majka) and the
current effort to make GNAT/DOS live again.

Stanley Allen
mailto:[log in to unmask]