[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Reasons for withdrawal or not



On Tue, Dec 09, 2003 at 06:38:34PM -0000, David Rush wrote:
> 
> 1) The collection hierarchy is woefully incomplete[2], even w/rt adequately
>   classifying the collections which are concretely specified in this SRFI

Of course it is.  A concrete collections SRFI is going to follow this 
one, probably authored by Mr. Campbell.  This SRFI was more about naming 
and interoperability than specifying the useful set of real collections.

> 
> 2) There is too much machinery required for compliance.

> I think that only *requiring* the implementation of the 'leaf'
> routines would cut right through these two difficulties; while leaving
> open the possibility of fixing some of the more difficult issues (a
> complete collections hierarchy that doesn't force compilers to break
> type hygiene, namespace management for leaf collection types) later.

I would disagree for the most part.  An implementation which doesn't 
allow one to write (list-add-from <some-ls> <some-vector>) wouldn't be 
that useful.  Agnostic collections use is a major goal of the SRFI.  

At any rate, the SRFI was 'finalized' late last month.  Its not 
officially finalized because Taylor hasn't had the free time to finish 
the reference implementation.

> 
> [2] the most complete and consistent collection hierarchy of which I am 
> aware
>    is the one in Smalltalk.

Smalltalk was heavily consulted.  This SRFI is in fact quite similar in 
what it covers.  What we don't offer (yet) are arrays, which are 
a natural subset of sequences for the one dimensional case, or of bags 
for a more general solution ala SRFI-48(?).  

	Scott

Attachment: pgpc4LgmjM9eg.pgp
Description: PGP signature