1983

Harris

Development of the Harris Word Processing System was not going all that well.  The software was mostly completed and undergoing testing.  The biggest problem was that it was extremely slow.  I had suspected this would be a problem, mainly because of the Harris Programming Language.  Dr. Haynam, my once professor at the University of Florida who developed the HPL language, told us developers “don’t worry about code size and performance, it could be optimized afterwards.”  Well, that was not turning out to be true.  The HPL complier produced bloated assembly language code which then ran on the 80186 processor.  Having considerable experience programming in assembly language from my years in the Harris Composition Systems division, they asked me look into it.  They had a sophisticated Intel development system that was hooked up to the 80186 allowing the code to be monitored, stopped, stopped and make detail time measurements.  I as identified bottle necks, I informed Dr. Hayman and he would make some adjustments to the code the HPL compiler produced although he didn’t seem all that interested in hearing it from me, just some programmer that didn’t even have a degree. 

The single biggest problem turned out to be what I had suspected from the start, string and text processing. With no text processing functions built into HPL, the programmers had to write functions to do simple tasks like search a string of text for a specific character or string of characters. Such functions are easy enough to write in a few lines of HPL code turning into fifty lines or more of assembly code which what ran on the system.   The code became bloated, more memory to the system was added increasing the cost.

The thing is, the Intel 80186 processor we were using had built in string instructions, what HPL did in fifty lines, could actually be done with a single 80186 instruction and execute in a fraction of the time.  To prove my point I took one of the modules I had written and patched into the code the single 80186 instruction in place of the fifty lines of code and tested it.  The module ran about 200% faster.  The problem was that was just for one HPL function, there were hundreds of such functions. To make the matter worse, since there were five teams developing different portions of the system, each team created their own functions that did basically the same thing but were coded differently.  I suggested a set of perhaps a dozen text functions be created, based on the 80186 instructions, and all groups use the set.  Well that didn’t go over to well, it would take weeks to rewrite the each group’s code.

By this time the newer Intel 80286 had been out for some time so the hardware guys suggested switching to it as it was faster and could handle more memory, but it would take some time to develop a new main processing board, but they could get it done in month or so.  I begin to see the writing on the wall as they say, the project was not going to succeed.

Reports were generated, suggestions made and then management decided to look at buying another company’s word processing system.  At first this wasn’t for everyone to know, but being on the architecture team I found out as they flew me and a couple of the other architects to New York City and to Atlanta to look at other companies’ products, even going so far as examining their code.

In New York, we began the day arriving early at the airport which the company we were visiting had a limo waiting to pick us up. It was a stretch limo for there were six of us from Harris, three from business and management, me and two of the other system architects.  I had never been to New York before, it was nice going though town in a limo to the company’s office.   Upon arriving we were greeted and then had an introduction meeting that lasted less than an hour then we broke up into two teams, the business and management stayed in the nice conference room while me and the other system architects were shown to windowless conference room where a guy from their team gave us an overview of their system and design.  Then he left us to “review the code”, lunch would be brought in later he said upon leaving.  On the conference room table were six or eight binders, each several inches thick, of 11x14 of computer print outs of the code for me and the other architects to review.  Late afternoon we were called to the management meeting and gave a report of what we found.  I’m not sure what they expected, three guys giving a few hours to review tens of thousands of lines of code.  At the end of the workday a couple of managers their company took us all, again in the stretch limo, to a nice restaurant for dinner then after dinner they took us to the play Annie. I had never been to anything more than a high school play before, so I thought it was quite spectacular.  After the play the limo was there to take us to the hotel, the only problem it was no longer the stretch limo, it was just a large, black Town car.  The three managers got the back seat while the three of us architects squeezed in the front seat next to the driver.  I forget who sat on whose lap on the way to the hotel.  The next morning we took a couple of taxies to the airport and returned back to Melbourne.

In the end Harris gave up on creating their own word processing system and acquired Lanier Business Products, Inc., a dictation, word processing and computer company based in Atlanta, Georgia.  But that was not the end of the story, what finally put an end to all such dedicated word processing systems was the IBM PC and WordStar.  

IBM PC/XT