Re: Capabilities interact nicely with Substructural Types and Reactivity

On Fri, September 13, 2013 00:38, David Barbour wrote:
> On Thu, Sep 12, 2013 at 2:09 PM, Rob Meijer <rmeijer-qWit8jRvyhVmR6Xm/[email protected]> wrote:
>
>> On Thu, September 12, 2013 18:21, David Barbour wrote:
>> > I wasn't speaking of creating any *particular class*, but rather the
>> > *ambient authority to `new` in general*.
>> >  [..]
>> > Private constructors have no effect on this at all.
>> >
>>
>> On itself it might not, but when combined with other language features,
>> it
>> could:
>>
>
>>
>> http://minorfs.wordpress.com/2013/01/18/raiicap-pattern-injected-singleton-alternative-for-c/
>
>
> What I described (ambient authority to create state and identity, i.e.
> *any* new objects)
> and what everyone else here seems to be hearing (ambient authority to
> create *specific* new objects) are very different issues. Unfortunately,
> like many writers, I'm in a bad position to understand why this
> distinction
> isn't as super-obvious and clear to everyone else as it is to me.
>
> Could you explain how you think singletons are related?
>

Singletons are not directly related, other than that I wrote up this blog
post on this pattern in order to offer a more DI solution (for C++) to
resource management situations where Singletons are often used by C++
developers.

The pattern basically asks the creator of the class Foo to define its
constructors using a constructor argument of type const reference to
raiicap<Foo>. Given that raiicap<Foo> has a private constructor 'and'
a friend definition for the program's main, only main can ambiently claim
the authority to construct a Foo. This const reference to raiicap<Foo> can
than be delegated to other parts of the program that need to be able to
construct Foo objects.

I understand that in C++, the 'any' case can't be solved like this, but
there is no reason a construct like this could not be used in a C++
inspired object capability language that would make the const reference to
raiicap<Foo> constructor argument part of the language specifications for
constructors.

Hope I'm making some sense here,

Rob