[ih] "The Great Debate"

[Craig Partridge]

On Sun, Apr 26, 2026 at 10:04 AM John Day via Internet-history <
internet-history at elists.isoc.org> wrote:

> I always considered Marshall’s ISODE implementation as a classical example
> of of how not to implement OSI. It was so bad, I always considered it more
> an attempt to trash OSI than promote it. We had implementations orders of
> magnitude smaller and faster.
>

I think this may be a misunderstanding of what Marshall was likely doing (I
will say, my primary interactions with Marshall's code were for other
projects and I don't recall ISODE code well).

Marshall had (and presumably still has) a particular software engineering
talent, which is producing working implementations of complex standards (or
APIs or defined systems) swiftly.  He used a particular implementation
style -- he was a devotee of "a function does one thing only", which meant
his functions were short and easy to test, but it meant that software often
made function calls dozens of levels deep -- and sometimes had redundant
loops.  Once upon a time Marshall and Van Jacobson collaborated on a
project -- Marshall got the initial code working and then Van improved it
-- Van cheerfully observed that Marshall had produced, if I remember right,
an O(n^3) solution to an O(n) problem due to nesting -- but also had gotten
the code up and testable in record time.  That was Marshall's genius at the
time (may well be now).

I loved Marshall's code -- it came up promptly, with very few bugs -- in a
world were being able to have running code and see what it was like, that
was a huge win.  I remember discussing this pattern with someone at the
wonderful Army Research Lab team [the folks who made so many contributions
to public Unix apps and he observed they had a person like that on their
staff and that person was invaluable.  "[He] gets it up and working and
then if it was a good idea, we send someone else in to refine the code."

Craig

-- 
*****
Craig Partridge's email account for professional society activities and
mailing lists.