[Uim] Proposal for SigScheme

[YamaKen]

At Fri, 22 Jul 2005 11:37:41 +0900,
mover at hct.zaq.ne.jp wrote:
> > Sometimes Scheme code consumes much more memory than C,
> > so that depends on the situation.
> >
> > In addition, development cost is also important. If
> > writing in Scheme is considerably easy, Scheme code sould
> > be used. Rewriting with C later is always possible.
> Now I decided to write SRFI-1 list procedure library in C for
> considering memory consumption, you pointed out, and good
> C interface.
> 
> Development cost is not the problem (Maybe 1 or 2 days are
> needed for getting things done). But this feature should be
> configurable, because I intend SigScheme is not only for uim.
> It aims to be embedded in various programs:-)

Did your words "development cost" contain writing automatic test
suites and ensuring ACCURATE specification well enough? I doubt
its easiness.

I think that implementing the multiple values and proper tail
recursion features to port the widely-used SLIB library to
SigScheme should take precedence to get SRFI features as first
step.

Although the C version of SRFI features are important for
SigScheme, it's only for performance and optional. But the two
core R5RS features are REQUIRED and important. I strongly
recommend developing them before your own SRFI implementations.

-------------------------------
YamaKen  yamaken at bp.iij4u.or.jp