Re: Ada GC

Henry Baker says, responding to Bob Duff


"> Some have pointed out that is difficult to *require* GC in a language
> standard, because it's difficult to *formally* specify what it means
> to have GC.


It's no harder to formally specify what it means to 'have GC' than it
is to specify what it means to be 'real-time'. Both storage & time
are resources which have been quantified in a measurable manner.
Since Ada/83/95 claims to be good for 'real-time', then I presume that
the Ada specifications make clear what the bounds for the various
primitive operations are, and which of these bounds are O(1), O(log
n), O(n), etc. (Ada doesn't do this? Pity. Then I guess Ada can't
be used for real-time systems. :-)"


Indeed, you can't really specify what it means to be real-time. The
best you can do is to require some documentation, which is rather
weak. That's what the Ada 95 RT metrics are about.


However, the idea that the failure to specify this means that Ada
can't be used for real-time systems is of course, as Henry's smiley
indicates, bogus. The analog would be to say that since Ada does not
specify GC, it is impossible to implement a version of Ada using GC,
which is also bogus.


One thing that is interesting here is that people will start to use
the Ada-Java product and get a taste of Ada-with-GC. Maybe they will
like it so much that ....
--

Return to the comp.compilers page.
Search the comp.compilers archives again.