On 11/10/2012 07:55 PM, Arthur A. Gleckler wrote:
> Still, it would be a disservice to provide it in an implementation that
> didn't support tail recursion, for example, since that's a basic feature
> of the language and many programs won't work without it.

Kawa does some non-trivial compile-time analysis that eliminates
many tail calls - including of course plain tail-recursion, and
some state-machine-style mutual tail-recursion.  So most Scheme
programs won't have a problem.  However, some styles/idioms will
have problems.

> It would be
> better to agree on a feature identifier that such languages can support.
> This can be outside the spec, for example in an SRFI.

One question to consider: What is the purpose of a feature identifier?
Specifically, when would it be useful to cond-expand on r7rs?

It might be useful to cond-expand on full-tailcalls, since one might
use different algorithms depending on the answer.
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/