Re: Standard Queue status

[Stephen Leake <stephen.a.leake.1@gsfc.nasa.gov>]

Stephen Leake <stephen.a.leake.1@gsfc.nasa.gov> writes:

> I'd like to suggest a slightly different approach. 

And here are my votes on these criteria.

> user must be able to get a list package with a single
> instantiation. 

yes.

> lists must be safe (ie no dangling pointers, etc) against _any_ list
> or iterator operation

No. There should be an "Unchecked_Lists" package that is _fast_, and
perhaps another "Checked" version built on top of it.

> lists must be efficient enough for hard real-time use

No. Hard real-time doesn't do dynamic allocation.

> lists must be safe in a multitasking environment

No. Similar to "Unchecked" above.

> lists must not be a tagged type

I could go either way. Making it tagged allows deep copy on a
non-limited list, so I lean that way.

> lists must be safe for assignment (always do deep copy, or don't
> allow assignment).

Yes.

> list elements must not be private

Hmm, I meant "limited private", not "private". And all my others are
stated positively. So let's replace this with:

"List elements must be limited private".

Yes. They force the user to provide a Copy routine, allowing
multi-layered deep copies.

> lists must support elements of any Ada type (private, limited,
> tagged, indefinite)

Yes. But I could relax this one if it really is a hassle.

-- 
-- Stephe