Re: binding thickness indicator, was Re: GNAT, OS/2, Libraries

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

tmoran@bix.com wrote:
: I propose we use something analogous to house insulation 'R-values'
: for indicating the thickness of a binding.  R1 would be the thinnest
: possible, a la single pane window glass, like Tore's example of
: interfacing to a C routine.  R50 would indicate such thickness and
: insulation that, say, code using an R50 windowing GUI binding would be
: portable across Mac, Windows, OS/2, and X-Windows.  R5 might be the
: Ada flavor he mentions (type and range checked parameters, raise
: exception instead of returning success/fail flag, etc) , but pretty
: much still tied to the particular design, style, and widget offerings
: of a particular system.  Clearly not an exact measurement, but
: somewhat more specific than thick/thin.

A good way of measuring the thickness is the size of the Ada package
bodies and supporting code relative to the Ada package specifications.
If almost everything can be done in the Ada package spec with 
type, data, constant, and interfaced subprogram declarations, the binding
is very thin.  The more code that's needed in the bodies, the thicker
the binding.  For example X11Ada, which I consider to be a thin binding,
contains 15K (non-blank, non-comment) lines in package specs and 6K
lines in package bodies.  72% of the code is in the specs.

- Mitch Gart