[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
On Sat, 2010-03-06 at 13:11 -0500, Aaron W. Hsu wrote:
> On Fri, 05 Mar 2010 20:57:04 -0500, Derick Eddington
> > I'm not sure what to do...
> Just get rid of the special character encoding behavior. Generally
> speaking, I don't like the idea of tying things like this to an operating
> system, and you're basically doing this.
That's certainly what's being done. That's the purpose of SRFI 103.
It's a real issue, currently causing problems, which needs to be
addressed. If the SRFI was not tied to specific OSs, it would not be
appropriate to assume anything about what a file name is or if there's
even a file system. Scheme systems are free to do something else on OSs
other than Windoze and unixes, and SRFI 103 should be trashed when
Windoze is finally dead. The only aspect which is affected is
library-file locating, all other aspects of using a Scheme system are
orthogonal, so changing to some other standard for locating libraries
will be cleanly doable.
> There's not a good way to do
> this, I think. It would be better to just let people deal with it on their
Requiring people to continually deal with such portability hassles is
not going to help attract more people. (But it might help increase rage
against Windoze, which I'm okay with.)
> by either restricting the names
Restricting library names because of Windoze's flaws is not acceptable.
I'm definitely not going to do it myself. See my recent message at the
SRFI 103 mailing list for more.
> or loading the files manually.
> There's nothing wrong with loading libraries explicitly, rather than
> having them implicitly loaded when imported.
I disagree and cannot support that. That's imperative style and not
compatible with Scheme systems which don't have some outside context
from which to execute the loading. R6RS libraries are supposed to be
usable as the fundamental source-code unit, which necessitates
auto-loading of imports. I think being compatible with systems which
have an outside context (e.g. a traditional top-level) is good, but SRFI
103 must not discourage the use of all possible library names (because
I'm not going to support corrupting symbols' usefulness) in systems
which do not have that.