Jump to content
 

The Story of Mel


Orunmila

4,279 views

 Share

The story of Mel Kaye is perhaps my favorite bit of Lore. Mel is often referred to as a "Real Programmer", of course a parody of the famous essay Real Men Don't Eat Quiche by Bruce Feirstein which led to the publication of a precious tongue-in-cheek book on stereotypes about masculinity, which in turn led Ed Post to speculate about the properties of the "Real Programmer" in his essay Real Programmers Don't use Pascal, but I digress.

The story itself is famous enough to have it's own Wikipedia page. Even though this was written 36 years ago back in 1983 it still resonates well. The story shows us, by the mistakes of others, that doing something in an impressively complex way may very well demonstrate impressive skill and truly awe onlookers, but it is really not the best way to do things in the real world where our code needs to be maintained later.

A good friend once told me that Genius is like a fertile field. However much it's potential to grow fantastic crops, it is also capable of growing the most impressive weeds if left untended, as Mel has shown us first hand here.

Without further ado, here it is for your reading pleasure:

The story of Mel

A recent article devoted to the *macho* side of programming
     made the bald and unvarnished statement:
                Real Programmers write in Fortran.

     Maybe they do now,
     in this decadent era of
     Lite beer, hand calculators and "user-friendly" software
     but back in the Good Old Days,
     when the term "software" sounded funny
     and Real Computers were made out of drums and vacuum tubes,
     Real Programmers wrote in machine code.
     Not Fortran. Not RATFOR. Not, even, assembly language.
     Machine Code.
     Raw, unadorned, inscrutable hexadecimal numbers.
     Directly.

     Lest a whole new generation of programmers
     grow up in ignorance of this glorious past,
     I feel duty-bound to describe,
     as best I can through the generation gap,
     how a Real Programmer wrote code.
     I'll call him Mel,
     because that was his name.

     I first met Mel when I went to work for Royal McBee Computer Corp.,
     a now-defunct subsidiary of the typewriter company.
     The firm manufactured the LGP-30,
     a small, cheap (by the standards of the day)
     drum-memory computer,
     and had just started to manufacture
     the RPC-4000, a much-improved,
     bigger, better, faster -- drum-memory computer.
     Cores cost too much,
     and weren't here to stay, anyway.
     (That's why you haven't heard of the company, or the computer.)

     I had been hired to write a Fortran compiler
     for this new marvel and Mel was my guide to its wonders.
     Mel didn't approve of compilers.

     "If a program can't rewrite its own code,"
     he asked, "what good is it?"

     Mel had written,
     in hexadecimal,
     the most popular computer program the company owned.
     It ran on the LGP-30
     and played blackjack with potential customers
     at computer shows.
     Its effect was always dramatic.
     The LGP-30 booth was packed at every show,
     and the IBM salesmen stood around
     talking to each other.
     Whether or not this actually sold computers
     was a question we never discussed.

     Mel's job was to re-write
     the blackjack program for the RPC-4000.
     (Port?  What does that mean?)
     The new computer had a one-plus-one
     addressing scheme,
     in which each machine instruction,
     in addition to the operation code
     and the address of the needed operand,
     had a second address that indicated where, on the revolving drum,
     the next instruction was located.
     In modern parlance,
     every single instruction was followed by a GO TO!
     Put *that* in Pascal's pipe and smoke it.

     Mel loved the RPC-4000
     because he could optimize his code:
     that is, locate instructions on the drum
     so that just as one finished its job,
     the next would be just arriving at the "read head"
     and available for immediate execution.
     There was a program to do that job,
     an "optimizing assembler",
     but Mel refused to use it.

     "You never know where its going to put things",
     he explained, "so you'd have to use separate constants".

     It was a long time before I understood that remark.
     Since Mel knew the numerical value
     of every operation code,
     and assigned his own drum addresses,
     every instruction he wrote could also be considered
     a numerical constant.
     He could pick up an earlier "add" instruction, say,
     and multiply by it,
     if it had the right numeric value.
     His code was not easy for someone else to modify.

     I compared Mel's hand-optimized programs
     with the same code massaged by the optimizing assembler program,
     and Mel's always ran faster.
     That was because the "top-down" method of program design
     hadn't been invented yet,
     and Mel wouldn't have used it anyway.
     He wrote the innermost parts of his program loops first,
     so they would get first choice
     of the optimum address locations on the drum.
     The optimizing assembler wasn't smart enough to do it that way.

     Mel never wrote time-delay loops, either,
     even when the balky Flexowriter
     required a delay between output characters to work right.
     He just located instructions on the drum
     so each successive one was just *past* the read head
     when it was needed;
     the drum had to execute another complete revolution
     to find the next instruction.
     He coined an unforgettable term for this procedure.
     Although "optimum" is an absolute term,
     like "unique", it became common verbal practice
     to make it relative:
     "not quite optimum" or "less optimum"
     or "not very optimum".
     Mel called the maximum time-delay locations
     the "most pessimum".

     After he finished the blackjack program
     and got it to run,
     ("Even the initializer is optimized",
     he said proudly)
     he got a Change Request from the sales department.
     The program used an elegant (optimized)
     random number generator
     to shuffle the "cards" and deal from the "deck",
     and some of the salesmen felt it was too fair,
     since sometimes the customers lost.
     They wanted Mel to modify the program
     so, at the setting of a sense switch on the console,
     they could change the odds and let the customer win.

     Mel balked.
     He felt this was patently dishonest,
     which it was,
     and that it impinged on his personal integrity as a programmer,
     which it did,
     so he refused to do it.
     The Head Salesman talked to Mel,
     as did the Big Boss and, at the boss's urging,
     a few Fellow Programmers.
     Mel finally gave in and wrote the code,
     but he got the test backwards,
     and, when the sense switch was turned on,
     the program would cheat, winning every time.
     Mel was delighted with this,
     claiming his subconscious was uncontrollably ethical,
     and adamantly refused to fix it.

     After Mel had left the company for greener pa$ture$,
     the Big Boss asked me to look at the code
     and see if I could find the test and reverse it.
     Somewhat reluctantly, I agreed to look.
     Tracking Mel's code was a real adventure.

     I have often felt that programming is an art form,
     whose real value can only be appreciated
     by another versed in the same arcane art;
     there are lovely gems and brilliant coups
     hidden from human view and admiration, sometimes forever,
     by the very nature of the process.
     You can learn a lot about an individual
     just by reading through his code,
     even in hexadecimal.
     Mel was, I think, an unsung genius.

     Perhaps my greatest shock came
     when I found an innocent loop that had no test in it.
     No test. *None*.
     Common sense said it had to be a closed loop,
     where the program would circle, forever, endlessly.
     Program control passed right through it, however,
     and safely out the other side.
     It took me two weeks to figure it out.

     The RPC-4000 computer had a really modern facility
     called an index register.
     It allowed the programmer to write a program loop
     that used an indexed instruction inside;
     each time through,
     the number in the index register
     was added to the address of that instruction,
     so it would refer
     to the next datum in a series.
     He had only to increment the index register
     each time through.
     Mel never used it.

     Instead, he would pull the instruction into a machine register,
     add one to its address,
     and store it back.
     He would then execute the modified instruction
     right from the register.
     The loop was written so this additional execution time
     was taken into account --
     just as this instruction finished,
     the next one was right under the drum's read head,
     ready to go.
     But the loop had no test in it.

     The vital clue came when I noticed
     the index register bit,
     the bit that lay between the address
     and the operation code in the instruction word,
     was turned on--
     yet Mel never used the index register,
     leaving it zero all the time.
     When the light went on it nearly blinded me.

     He had located the data he was working on
     near the top of memory --
     the largest locations the instructions could address --
     so, after the last datum was handled,
     incrementing the instruction address
     would make it overflow.
     The carry would add one to the
     operation code, changing it to the next one in the instruction set:
     a jump instruction.
     Sure enough, the next program instruction was
     in address location zero,
     and the program went happily on its way.

     I haven't kept in touch with Mel,
     so I don't know if he ever gave in to the flood of
     change that has washed over programming techniques
     since those long-gone days.
     I like to think he didn't.
     In any event,
     I was impressed enough that I quit looking for the
     offending test,
     telling the Big Boss I couldn't find it.
     He didn't seem surprised.

     When I left the company,
     the blackjack program would still cheat
     if you turned on the right sense switch,
     and I think that's how it should be.
     I didn't feel comfortable
     hacking up the code of a Real Programmer."


         -- Source: usenet: utastro!nather, May 21, 1983.

 

  • Haha 1
 Share

2 Comments


Recommended Comments

  • Member

This brings back memories.  As I developed as a programmer, the role of Mel (and his suitability as a role model) has also shifted.  Of course we all strive to be Mel when we start out.  The idea that we have such in-depth knowledge of our system that we can create the tightest possible code and take advantage of the most esoteric of "features" seems like the epitome of super programmers.  Of course as I matured, I learned that this knowledge was incredibly powerful and certainly the ability to use it was an asset but like the old saw, "with great power comes great responsibility".  Today, I think the best analogy would be "absolute power corrupts absolutely". I would love to manage a software team with the knowledge and experience of Mel, but I would hate to manage Mel, because in the end, he would cost me a huge amount in future maintenance costs as only another Mel + lots of time could possibly support his work and no program ever survives contact with the users so the inevitable new feature would be a disaster.

 

That aside, be sure to read the detailed explanations.  They are as fascinating as the original story.

https://jamesseibel.com/the-story-of-mel/

https://medium.freecodecamp.org/macho-programmers-drum-memory-and-a-forensic-analysis-of-1960s-machine-code-6c5da6a40244

 

  • Wow 1
Link to comment
Guest
Add a comment...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...