Re: Quality of Thin Bindings (was: Programming X Windows...)

[mg@dsd.camb.inmet.com (Mitch Gart)]

Larry Kilgallen (kilgallen@eisner.decus.org) wrote:

: I have looked at the Win32 bindings which come with ObjectAda 7.0
: and it seems to me they rely excessively on a single integer type.

The problem is that Win32Ada doesn't represent an independent design 
of an API, it is a translation of a Microsoft API into Ada.  If 
Microsoft uses universal integer types everywhere, so must the binding.

Consider a simple example:

#define MASK1 1
#define MASK2 2
#define MASK3 4

In Ada is this a series of integer constants or an enumeration?  It's
tempting to say it should be an enumeration but then what about functions
in the binding that take integer parameters and return integer results?
Some of those functions may in fact take mask values or return mask results
from this enumeration, but how can the person writing the binding be sure
of the intentions of the original authors of the C software?  This kind of
question is one reason why Win32Ada is at such a direct one-for-one level
with the original C software.  

I and others have said many times that it would be good to think of a binding
like Win32Ada as a low level layer upon which other higher level layers could
be built.  I would encourage people to use Win32Ada as a base level and build
higher level Ada bindings to specific small parts of the Win32 API, for example
MAPI or OLE or ODBC or Video For Windows.

(snip)
: I don't know what ACT might be doing to strengthen Starlet
(snip)

I'm guessing that the people who invented Starlet had two advantages:

- VMS and Starlet were both Digital products so presumably the people working
  on Starlet were able to easily talk to the VMS people 

- DEC was at one time funding Starlet development and saw Ada as a money-making
  business area.

Mitch Gart