Cameron Laird's personal notes on Stackless Python

Stackless Python is one of several interesting alternative implementations of the Python programming language.

Stackless has its own Web home.

Stephen Figgins does an excellent jobs of describing the state-of-the-implementation in his February 2002 "Stackless Reincarnate".

Christian, on 31 December 2001: "I hope to make it simpler, split apart Stackless and optimization, and continuations are no longer my primary target, but built-in microthreads. . . . What Python needs is a secure mechanism to switch frame changes at certain times." On 20 January 2003: "Stackless focuses on tasklets and channels, now."

Documents

Key Stackless writings include:

Christian's original paper for IPC8.
Guido's May 1999 reaction to continuations and Stackless Python.
Gordon McMillan's six-page introduction to Stackless Python.
Courageous' explanation of why stacklessness is wonderful, and in particular its performance advantages in terms of threading.
Timbot's magisterial 1994 demonstration of Python co-routines in terms of threads.
Stephen Figgins explains Stackless for O'Reilly Network; not long after, in October 2000, they published my "Introduction to Stackless Python", "Programming Stackless Python", and "(Stackless) Python Roadmap".
I also wrote "Is Stackless Python for You?"
[PEPs]
David Mertz' interview with Christian Tismer and John Max Skaller
Reynolds "The Discoveries of Continuations"

Programming constructs

[pointers within documents above]

Continuations

In 1994 (? earlier?), Steven Majewski was designing an implementation of Python that supported continuations. As part of that research, he reported back to Guido, Tim, and others this summary of continuation theory.

[Schemers emphasize that language's two decades of experience with continuations.] [But compare with Smalltalk control flow ...]

Co-routines

In reflections on his understanding of control constructs as of 1994, Steven Majewski wrote me that

Knuth made the distinction between full coroutines and what he called "semi-coroutines" . Icon has both co-expressions and coroutines, but although the language manual made it clear when you would use one or the other, it didn't stress the foundational difference. ( co-expressions or generators in Icon are Knuth's semi-coroutines. )
Coroutines are fully symmetrical -- control is explicitly passed from one routine to another. Semi-coroutines are asymmetrical -- they always return to their caller ( with a suspended state. ).

Generators

People

Along with the roll-call that leads off my "Introduction to Stackless Python", there are several other contributions to recognize:

Steven D. Majewski summarizes that Scheme's experience teaches us that "Continuations are sort of the functional equivalent of goto's (plus C's longjump with a little of Intercal's COMEFROM thrown in! ;-)." Exceptions are a similarly ambiguous area, where it's still hard engineering to make the right design choices. He seems generally to favor giving Python programmers all the good stuff--generators, co-routines, ...--that continuations make practical, but sternly discouraging them from writing directly in terms of continuations.
Majewski came close to getting Icon- and Scheme- (and eventually Simula- and Forth-)inspired continuations in Python ?.? in ?. He found the Scheme literature only moderately helpful. It was LISPly syntax-free, of course, and largely focused on compilation challenges. However, in his words, "Scheme did have a big role in generalizing from coroutines and generators to the idea that if you had continuations, you could build any kind of control sequencing."
...

Prospects

"Some sort of resolution to Stackless Python seems likely for 2.1."

[Explain Stackless and Palm. Also ...]

Cameron Laird's personal notes on Stackless Python/claird@phaseit.net