Re: [Sml-implementers] Sensible Structure Sharing for Standard ML

[Derek R. D. <dr...@cs...>]

No, Leaf, I think Martin is saying that "structure sharing is
transitive, etc." for *his* rule that he's proposing, not for ours. 
Likewise, he's saying that *his* rule is less permissive than the
Definition, not ours.  I'm still trying to decipher what Martin's rule
does, but it seems quite unrelated to our proposal, so I'm not sure what
Martin means by saying it is "based on" our proposal, maybe that it is
based on the alternative proposal idea described on part 4 of our
Discussion section.  Perhaps, Martin, you could explain what it does in
layman's terms?

In particular, I am still somewhat mystified by the requirement that
\rea(E_1) = ... = \rea(E_n).  This would seem to force the signatures of
all structures being shared to be pretty much identical.  For instance,
it would seem to force all the structures to have exactly the same set
of value components.  Why?  I mean it's an interesting idea, but perhaps
overly restrictive.  Am I wrong about my interpretation of your rule,
Martin?  Examples of how it works (of the sort we gave in our design
note) would help.

Thanks!
Derek

Leaf Eames Petersen wrote:
> 
> Martin -
> 
> Thanks for taking the time to look into this.  I have a couple
> questions about one of your comments though:
> 
> >    1. Structure sharing is transitive, reflexive, and commutative,
> 
> I don't believe that this is true.  Consider the following spec
> 
> structure A : sig type t end
> structure B : sig type s end
> structure C : sig type t end
>  sharing A = B
>  sharing B = C
> 
> This does not introduce the same equations as "sharing A = C", which
> is what I would expect you to mean by transitive.  Perhaps you
> mean that structure sharing is transitive for the subset of the
> type names common to all the signatures in question?
> 
> >       thus the rule is less permissive than SML'97 in this respect;
> >       see The Def. page 85.
> 
> Our intention was to make our version strictly more permissive in the
> sense that all structure sharings accepted by the Definition would
> still be accepted by our proposal, and have the same effect.  Are you
> claiming that this is not the case?  If so, can you give an example of
> something that breaks?
> 
> Cheers,
> Leaf
> 
> _______________________________________________
> Sml-implementers mailing list
> Sml...@li...
> https://lists.sourceforge.net/lists/listinfo/sml-implementers