(j3.2006) derived types with type parameters are different

[Jim Giles]

----- Original Message ----- 
From: "Michael Ingrassia" <michaeli at ranma.sfbay.sun.com>
To: <j3 at j3-fortran.org>
Sent: Tuesday, July 29, 2008 2:46 PM
Subject: Re: (j3.2006) derived types with type parameters are different


> >1) Too drastic: Disallow parameters on SEQUENCE types
> 
> But this is equivalent to 
> Disallow SEQUENCE on parameterized types
> which sounds reasonable to me.
> 
> If it causes significant loss of functionality, what exactly is the
> functionality that we lose?

When inheritence becomes common I will be using SEQUENCE
on most types in order to guarantee they won't be extended.  The
committee removed the EXTENSIBLE keyword at the last minute
(and after the public review, as I recall).  Other than SEQUENCE
or BIND(C), I can't think of anything that prohibite type extension.

Sure, it's possible that all people that extend a data type will 
treat it as a black-box and not rely upon any internal properties.
Sure.  Boy scout programming.  But, I want the freedom to 
alter the innards of a data type without having to warn all
possible users of the fact.

Now, whether I'll evern need type parameters or not is a 
different question.  I still am not quite sure of all the issues (and
probably won't be until I can play with an implementation of
them).

J. Giles