Re: Let's gather some evidence!
On Thu, Sep 10, 2009 at 6:33 PM, Sylvain Pion
<Sylvain.Pion@xxxxxxxxxxxxxxx> wrote:
> Hi John & P1788,
> [...]
> The argument against NaI in principle is that it adds a cost in terms of the
> complexity of the specification. Suddenly it introduces special cases
> everywhere.
> Look at floating-point : it is complex to use (you need to read Goldberg),
> because of special cases and rounding issues. With IA, we have a chance
> to define a model of computation with approximations of reals which is
> greatly simpler, and it should be a selling point.
Indeed, the value of simplicity and regularity cannot be overstated.
Already, IA has a reputation of being specialist-only friendly -- when
in fact, we want
more people, e.g. non-specialists, to adopt the IA technology.
If the standard specification we come up with has a complicated conceptual and
usage model, then we have succeeded in cimenting forever a perception that was
probably not justified in the first place.
The danger of looking for the complete model, the silver bullet is that it is
so complicated that everybody agrees it has everything and nobody wants to
use it. We would have failed.
-- Gaby