Print

Print


Tucker Taft wrote:
>
> On the other hand, we have not worried much about validatability.
> We are trying to provide a useful tool.  We are using a validated
> Ada 95 front end, but we have not focused on producing a validated
> byte-code generator, but rather on a byte-code generator that is
> most compatible with the Java platform.  Someday we may achieve
> validation on this target, but probably only if a customer believes
> this is really necessary for the tool to be useful.

I think this is a reasonable approach. However, if Intermetrics did
decide to go for validation, they'd probably get a fair amount of
support from the AVO, etc. to make this happen -- even if some tests
had to be waived.

On the other hand, I get the feeling that some small developer, not
well known in the "Ada community," would have a much harder time
getting through the validation process with an Ada compiler targeted
to an 8-bit processor, if that developer decided to leave out certain
features because they were difficult to implement on that target. There
seems to be a fair amount of suspicion that, if an Ada feature is not
available, it must have been left out for a "bad" reason.

> This is certainly possible.  However, in terms of volume of developers
> and expected return on investment, Java happens to be a much easier bet
> to make than does perhaps some particular 8-bit architecture.  Although
> there are billions of 8051's out there, there are probably only 1000's
> of developers, whereas there are 100,000's of Java developers already,
> with the numbers expected to increase steadily.

No argument there. However, "1000's" isn't a bad market - if you have a
shot at a decent market share. I wouldn't claim that anyone is going to
get filthy rich building for small targets, but they might make a profit.
And, consider what happens as these thousands of developers migrate to
more powerful architectures. It's not completely insane to treat the
tiny processor market as a loss leader - if you have the 16-bit and
32-bit compilers available for a subsequent migration, and you can stay
in business long enough to survive the up-front losses.

> But numbers of developers isn't everything.  Ken is right that
> we did the Ada/Java work mostly because we found it new and interesting
> and potentially profitable.  We didn't do a market survey first.

And, I'd guess that this is the reason for the existence of various
other compilers - say, for OS/2 (who uses THAT? :). Nothing wrong
with this. Unfortunately, 8-bit processors are not particularly sexy.
So, if I ever want to work on a project where we actually get to 100%
Ada use, I either have to

  a. wean all subcontractors from tiny processors (fat chance),
  b. find some way to convince a vendor to take an interest
     (without paying for the development each time), or
  c. build the tools myself (tempting, but not a good long-term solution).

> -Tucker Taft   [log in to unmask]

--
LMTAS - "Our Brand Means Quality"
For more info, see http://www.lmtas.com or http://www.lmco.com