[PRL] Type question...

[Matthias Felleisen]

On Oct 8, 2007, at 12:18 PM, Mitchell Wand wrote:

> OTOH, the Erlang community seems to be moving in the other  
> direction.  They have a notion of "success typing" that says  
> roughly, if the computation succeeds, then some property holds.  As  
> near as I can tell, this seems to confuse typing with flow  
> analysis. (Somebody here should probably write a oaper sorting all  
> this out.)


We definitely need to check this out, because our upcoming POPL paper  
on Typed Scheme is also using flow-oriented reasoning to supplement  
the logic of (union, sub, poly) types. Any pointers? -- Matthias



>
> --Mitch
>
> On 10/8/07, Riccardo Pucella <riccardo at ccs.neu.edu> wrote:
> >
> > Pascal used to assign types to this kind of program :-)
> > (not quite the (f f n) part) and I bet you can still get
> > it thru the C type checker but I assume you mean "sound"
> > type system (and so did everyone else who responded).
>
> If only because C has only a notion of a function type that does  
> not specify what the functions args and return types are. (At least  
> to a first approximation.)
>
> >
> > I wish regular programmers could distinguish these
> > two kinds of systems.
>
> I like to think of the distinction as between type systems that are  
> used to make sure you do not use values wrongly (sound type  
> systems), versus type systems that are used for conveying  
> representation information - and the latter is definitely what is  
> going on in C.
>
>
> >
> > Why don't you post this kind of question on a forum
> > where you can evaluate the reactions of regular
> > programmers. I wonder how much this idea of 'safety'
> > has sunk in (and what this says about PL education
> > at the undergraduate level. I used to cover Type
> > Soundness in "311" aka "660". I doubt it is still
> > mentioned.)
>
> I think Java has helped in that regard - at least for the average  
> programmer. They may not be able to put it into words that crisply,  
> but they have much more of a sense of soundness than the C  
> programmers (or Pascal programmers) of yore.
>
>
> Cheers,
> R
>
>
> _______________________________________________
> PRL mailing list
> PRL at lists.ccs.neu.edu
> https://lists.ccs.neu.edu/bin/listinfo/prl
>
>