function types syntax

[Neal Gafter]

On Wed, Jan 6, 2010 at 3:11 PM, Mikael Grev <grev at miginfocom.com> wrote:
> But what I mean is that the 95%-type solution, with a simple and likable rule set/syntax, is fine and
> preferable to the 99,7%-type solution with a more awkward syntax/rule set. Even as the end result.
> The last percentage points are bonus and can be handled on a feedback basis, if possible.

Exploring the obscure corner cases (i.e., combining a proposal with
other language features to see how it works) helps us to understand
where the issues are.  To someone not used to engaging in programming
language design, it may feel like we're focusing most of our attention
of the least likely use cases at the expense of the common use cases.
That isn't the case.  We're focusing on the more difficult use cases
to ensure that the specification is implementable in a simple and
general way, rather than a large set of hand-selected specific use
cases with hand-selected semantics.  We're ensuring that programs
written with the construct are easy to read and have intuitive results
even when used in nontrivial programs combined with other language
features.

The kind of heuristics you're suggesting for how we focus our
attention on use cases is much more appropriate to API design than
language design.  For a bit more discussion on the role of use cases
in language design, see
<http://gafter.blogspot.com/2006/09/failure-of-imagination-in-language_17.html>.

Cheers,
Neal