Re: [MIT-Scheme-devel] [commit fe044480] runtime: Make it nicer to (ge '(R3RS)).

[Matt Birkholz]

> From: Taylor R Campbell <address@hidden>
> Date: Mon, 6 May 2013 01:21:21 +0000
> 
> [...]
> Ugh...  The behaviour of the parser and unparser should not depend
> on whatever state the REPL user interface is in.  That leads to
> nonsense like IMAIL's IMAP client breaking when you ask the REPL to
> print numbers in hex.

IMAIL formats numbers with WRITE?!

> We need to eliminate this kind of thing, not add more of it.  Please
> back out this change.

Seriously?  I was boarding up a pit I fell into whilst exploring
packages with (parent #f).  You were not supposed to be affected.  I
did not change the behavior other than to retry before erroring.

Do you want another adventurer to have to hunt down and kill his
Scheme while it spews half-printed error messages out its console
forever, brother?

> Frankly I don't think environments are the right mechanism at all for
> configuring the parser.  We'd be much better off with a hash table of
> options (or a wt-tree so we can easily do set operations on them).

OK, but let's keep the top-level syntax and bindings together with
them.  They were designed to work together, yes?  And we needn't call
these tables "environments" nor even "packages".  At the risk of being
vague: how about "contexts"?  "Structures" ala s48?  Maybe "Taylors-
special-sauce" soothes the savage beast?

As long as it smells sweet... and doesn't go bat-crap crazy when I'm
trying to debug something.

For now the boards over that pit will stay in any branch of mine, but
if you insist... well, you have write access elsewhere.