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

Re: what is the problem?




Jeremy, as it turns out this form of "fear" isn't new. Dave Hanson (at the
time UofA, Tucson, now at MS Research) wrote a technical report with a
title like "Nested Scoping Considered Harmful". -- Matthias

  > Content-Type: text/plain; charset=us-ascii
  > Date: Tue, 18 Dec 2001 16:25:27 -0500 (EST)
  > From: jeremy@alum.mit.edu (Jeremy Hylton)
  > Cc: Paul Prescod <paul@prescod.net>, ll1-discuss@ai.mit.edu
  > Reply-To: jeremy@zope.com
  > Sender: owner-ll1-discuss@ai.mit.edu
  > Precedence: bulk
  > 
  > >>>>> "PG" == Paul Graham <paulgraham@yahoo.com> writes:
  > 
  >   PG> Perl has been getting Lispier, but it started further away from
  >   PG> Lisp than Python and Ruby.  As a first approximation, Python
  >   PG> seems to be an inadvertant reinvention of a 1970s
  >   PG> dynamically-scoped Lisp (except for s-expressions and thus
  >   PG> macros).  It seems as if these ideas came to Python indirectly
  >   PG> via Modula-3, which was based in part on Lisp.
  > 
  > Python wasn't dynamically scoped.  It didn't support nested scopes at
  > all.  Every function regardless of whether it was nested inside
  > another function had access to two namespaces -- the local namespace
  > and the global/builtin namespace.  The determination of the global
  > namespace is static; it's the globals where the function was defined.
  > 
  > I believe these odd scoping rules are the result of some bad
  > experiences with Pascal, where nesting functions was one of the few
  > structuring techniques for large programs.  So several people were
  > familiar with large Pascal programs that were very hard to understand
  > because they used many nested functions.  Python, then, wanted to
  > discourage this style of programming.
  > 
  > Jeremy
  >