[mercury-users] Circular lists

Richard A. O'Keefe ok at cs.rmit.edu.au
Thu Nov 27 22:18:27 AEDT 1997


	So what happens if you get stuck with an application Mercury
	does not give you the speed you need, and I am speaking of orders
	of magnitude!

Show me one where (decimal) orders of magnitude are involved.
(Binary orders of magnitude can easily show up in C with the
same code compiled by different compilers.)

	You advise those people to skip Mercury, and use C?

You made this up.  I never said any such thing.
I EXPLICITLY said that I am not into _forcing_ people to do anything
in particular.

I also made it plain that what I want is not to be put at risk *SILENTLY
BY DEFAULT*.  When there is a risk, TELL me, so I can make an INFORMED
judgement.

	I thought it was better to use the C interface to build a dirty
	module with high efficiency AND a clean interface.

Sure, and there isn't the slightest reason why such a thing should fail
the kind of checks I am asking for.

	I am VERY happy with this solution: I got efficiency, fast development,
	very few bugs, ... .

	You tell me I'm WRONG?

You are certainly very offensive to impute things to someone without any
justification and then attack them for it.  You don't seem to have read
what I actually wrote.

Let me try one last time.

In recent messages in this mailing list we have seen Mercury's C interface
to do things that are (a) dangerous, and (b) NOT essential to performance.
I have asked whether there could be or will be a tool that will warn about
dangerous things while letting the straightforward uses through.  I did not
ask about an option in the compiler to block such things, just a separate
tool to WARN about them.

If you want dangerous things to happen quietly,
if you don't want to make INFORMED choices about safety,
if you assume that speed and safety cannot be had together,
then we have nothing to say to each other.



More information about the users mailing list