[cap-talk] Last Call for ref_send API 1.0

[Jonathan S. Shapiro]

On Wed, 2007-04-04 at 18:09 -0700, Bill Frantz wrote:
> When I hear "reference implementation", I think of an implementation
> which isn't suited to real work, but may be useful as a worked example
> for standardizers or implementors.  For example, the first reference
> implementation of Ada was implemented in one of the Lisp family of
> languages.  It was updated after every standards meeting to include the
> results of that meeting.  As such, it was quite useful to the standards
> process, but as a friend who worked on it said, "As an interpreter
> running an interpreter, it's only suitable for real-time simulation of
> paper and pencil calculations."

[Completely off topic]

Damn it's a small world. The first validated Ada translation system was
built by Charlie Weatherall's group at Bell Labs just around the corner
from my office. It was also an interpreter, and it was "validated" in
the sense that it passed the validation suite.

Once having acquired the ego points of passing the suite, AT&T killed
the project because there was no way that system would ever be useful
for real code. The language was pretty complex, and there were just too
many corner cases for *any* team to cover, no matter how talented.

It would be many years before useful Ada compilers would emerge.

shap