Marcin 'Qrczak' Kowalczyk <xxxxxx@knm.org.pl> writes:

> Why no current implementation uses sparse arrays as the representation
> of (some) vectors? Or does any?

Maybe none has gotten around to it?

> If no implementation does a particular thing, then perhaps it's not
> a wise choice to do.

That's a fundamentally bogus argument.  If there's a reason it's
unwise, then let's hear it.  It needs to be a reason why it's *always*
unwise, not just why it's unwise in general-purpose Scheme systems.
(Though for the life of me I can't think of any reason there either.)

No Scheme implementation that I know of supports writing Scheme
programs that include identifiers written in the Devanagari script.
But that doesn't mean it's an unwise thing to do; it's just that there
aren't many Sanskrit users of Scheme yet.

> Leaving room for it only reduces programmer confidence and
> encourages them to rely on common practice instead of on standard
> guarantees.

"Reduces programmer confidence"?  Huh?  Programmer confidence in what?

It's you that seems to have mistaken common practice for a standard
guarantee.  Rewriting the standard so that it standardizes common
practice is not the solution.

Thomas