[plt-scheme] Are new Schemers supposed to be reading SRFIs?

From: Noel Welsh (noelwelsh at gmail.com)
Date: Fri May 11 03:08:36 EDT 2007

On 5/11/07, Eli Barzilay <eli at barzilay.org> wrote:
> Yes, but consider a different future: the new "srfi-300" list library
> comes out, *then* people start talking about the need for common names
> and obviously they want to choose "list-old" for srfi-1 and "list" for
> sefi-300.  At this point you're stuck.

This is a classic CS argument. Do we build a system that works for all
time, but is inconvenient, or build something that is convenient but
might break in the future?  The Right Way answer is the former, but I
prefer the later.  Here are two reasons:

 - Gaining users now is more valuable than gaining users in the
future, due to the same reason money now is more valuable than in the
future: the return the language gets from today's users via increased
libraries and visibility will pay off as greater usage in future.

 - PLT Scheme has to be prepared to (and does!) break compatibility
between releases.  If I wanted a language that wasn't on the cutting
edge I'd use Python.  The key differentiating feature of PLT Scheme is
that it incorporates the latest research, which necessitates revising
old decisions.

These kinds of arguments go against the classic CS culture but I think
they are compelling in a more realistic world view. YMMV.


Posted on the users mailing list.