Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Include language bindings? Re: Draft Standard Text, V02.1



> Subject: Re: Include language bindings? Re: Draft Standard Text, V02.1
> From: John Pryce <j.d.pryce@xxxxxxxxxxxx>
> Date: Tue, 16 Mar 2010 10:07:08 +0000
> To: stds-1788 <stds-1788@xxxxxxxxxxxxxxxxx>
> 
> P1788, Dan
> 
> On 15 Mar 2010, at 16:34, Dan Zuras Intervals wrote:
> >> That's a good question.
> >> 
> >> Looking at IEEE 754-1985, one sees that it was a decade or more
> >> after the hardware standard was in use before there were language
> >> bindings.  Can including language bindings help the process?
> > 
> > 	Emphatically yes.
> > 
> > 	It was more than 2 decades that 754 lived without
> > 	standard language bindings & during that time a lot
> > 	of compilers found 'imaginative' ways to interpret
> > 	the meaning of the seemingly most obvious things,
> > 	like '+' & '*'.
> > 
> > 	Some of those interpretations were just 'different'.
> > 
> > 	Some were flat out wrong...
> 
> Dan has long argued this point, and I am persuaded by him.
> We need language bindings. This is what the P1788 Subgroup
> "Expression Rearrangement" is really about. Can it wake from
> sleep and address this please?
> 
> This is tricky, because by the nature  of language bindings
> I think we, acting alone, can't create normative text on this.
> Or would be unwise to do so. That can only be done in
> collaboration with a relevant language standardization committee.
> If my health is spared to me I intend to join the Fortran
> committee when our work is near completion.
> 
> But if we produce a P1788 clause of _recommendations_ on this
> topic, they will carry a lot of weight. We should then be ready
> to revise it a year or two or three later in the light of
> experience in language committees.
> 
> Regards
> 
> John

	John is quite right.

	We will be unable to completely specify language bindings
	without the help of the various language standards
	committees.

	However, I feel we MUST do something more than just
	'recommend' bindings in the text of 1788.  We must word
	it in some way that implies that whatever bindings a
	given language committee chooses, it should be mandatory
	in all conforming instances of that language.

	I am unsure as to just HOW to make a 'shall' requirement
	that does not exactly specify how it is to be implemented.
	Perhaps those of you who are language experts may be able
	to suggest some approaches.

	For example, most of you speak of interval expressions in
	analogy to floating-point expressions.  We could, I suppose,
	ask (demand?) of the language committees that the language
	bindings for interval expressions be the same as for that
	language's floating-point expressions.

	But I see flaws in that approach:
	- The analogy is not exact & interval expressions have
	different meanings than their corresponding floating-point
	expressions.
	- Due to the reasons I've already outlined, many of the
	standardized languages that may be targets for our standard
	have ill defined floating-point expressions.  We would not
	want to infect 1788 with that flaw.
	- Interval expressions seem to need a bit more 'expressive
	power' than floating-point expressions.  In particular,
	the reverse or inner expressions need to have a way of
	being expressed.  I'm sure there's more.

	Still, clause 10 in 754-2007 manages to specify standardized
	language expression bindings without saying what they are.
	It is normative but optional.  We will need to make it
	mandatory.  IMHO, that specification is also flawed & has
	yet to be tested in the crucible of an actual language
	implementation.

	I hope we can do better.

	Yours,

				Dan