Chapter 7: Upgrading The System
A blank check
For the sixth time in nine years Almira and I moved our family, this time from Washington, DC, to Midwest City, Oklahoma, where SMS Al Polston and Lieutenant Jerry Shuster had been monitoring the still shaky AWN since June.
The three AFCS programmers on our start-up team had come to DC from Detachment 10, 1800th Support Squadron at Tinker, AFCS’s centralized computer programming group. Det. 10 would provide office space for my small group. Over the next two years Operating Location (OL-7), Headquarters AWS would grow from two to thirty people and because of its size be redesignated Detachment 7, Headquarters AWS.
The Tinker software was running as smoothly as the basic design would allow. Major program changes would wait until Det. 10 and OL-7 met and agreed on future design features.
The 6th Weather Squadron (Mobile) was asked to provide administrative support to OL-7 and I made a courtesy call on their commander, Lieutenant Colonel Elwyn (Mo) Moseley, the senior weather officer at Tinker AFB. Colonel Moseley enjoyed being weather service’s elder statesman and did not hide his displeasure from me with AWS’s decision to place my function outside of his unit. As he expressed his feelings with enthusiasm I nodded politely and said nothing. I thought it best that he didn’t know I had recommended the arrangement.
After I got my family settled in, Al and I got down to work. I expected an early call from Scotty at headquarters inviting me to come to AWS and discuss plans. By Friday I began to feel like the Beetle Bailey cartoon character General Halftrack, the commander of Camp Swampy who keeps waiting for a call from the Pentagon that never comes, so I called Scotty and asked when we would meet.
His response was a pleasant surprise: “Jack we sent you down there to decide what to tell-us-to-tell-you to do. You have my notes from our meetings in January but by now you’re the expert. We expect you to take this project from our early thinking into the future.” Looking back thirty years I can see how very young and naive I still was.
Scotty’s words were all the guidance I would ever need and more. He just handed me a blank check that I immediately filled in and cashed and I felt ten feet tall, inspired by all this permission. Al and I were now ready and eager to meet with Major Lloyd Anna’s people and get things moving.
Lloyd was the commander of Det. 10 and an easy person to get along with, but we would have our differences. Lloyd’s staff had been maintaining the growing number of AFCS’s communication computer systems and the two of us were now jointly responsible for upgrading the AWN.
His staff and I had both already agreed that the entire package must be rewritten, but we differed on how to go about it. As in any close relationship we entered with different priorities, protocols, and also different personalities and on occasion would not understand each other at all. Lloyd and I were, however, flexible enough to let the sparks fly and still maintain a mutual respect.
I was willing to answer to my own chain of command for my goofs, and the demarcation between Lloyd’s and my responsibilities within the system was clearly drawn. AWS would process incoming low-speed weather data: separate messages, determine data type, and package the data to fit AWS needs while Lloyd’s people would queue the traffic for output and guarantee its delivery. Because of the large number of different weather data types, each with their unique codes, AWS would eventually write more lines of code than AFCS.
Our main differences were philosophical, centering on program-control. Early on I recognized that Lloyd would come to see my group as cavalier, and sometimes even reckless. He would see us as too eager to put new code on-line before we gave it an AFCS-type test and acceptance shake-down. Over the next five years that I was associated with the AWN we never slowed down, but instead continued to code and implement a succession of improvements as fast as we were convinced they were safe and would improve the quality of our product or broaden the scope of our mission.
AFCS had a large staff to go along with their broader mission including a group whose only job was to test and accept new code. These tests were necessarily much more complicated than ours because they involved the total system. We, on the other hand, could write a sophisticated decoder for one data type and run a large quantity of live data through the logic off-line and determine almost immediately – yes or no – it works or it doesn’t work. We could identify and correct most errors or holes in our logic at our desks. The same code would be run at all three of the source 418s.
Again, because they did what they did – write communication software – AFCS had to bunch their updates into large packages and spend much more time designing than we had to before any of it could be programmed and implemented.
During the spring of 1965 while the AWN was still being developed, Det. 10 had put a large package of program changes (patches) into the AUTODIN system that didn’t work. In fact, they brought this world-wide system to a near halt and Det. 10 was forced to remove all of the patches from the system and regroup. Since AUTODIN served the entire Air Force, a storm of high-powered complaints followed that I’m sure were embarrassing. As a result, Det. 10 was even more cautious and conservative. Five years later, in August of 1970, AWS would suffer a similar embarrassment when they transferred their computer flight plan program from Suitland, Maryland, to Global Weather at Offutt.
Our scope of responsibility was narrower than Det. 10’s, it was limited to within AWS. If we implemented a program change that lost good data I would have to answer for it, primarily to Global. By coordinating proposed changes with Global they could let us know almost immediately if they saw a problem that we hadn’t detected and we could just as quickly switch back to the previous system, however this rarely happened.
When the AWN officially went on line on the 1st of July we had already met our primary commitment to the Air Staff, the intelligence community, the Strategic Air Command, and to Air Weather Service to deliver hourly aviation weather observations to Global within five minutes of origination.
Global comes on line
Sometime in late 1965, Offutt’s 418 came on-line and AWN data now was being delivered directly into Global’s main-frame computers – into their data base and analysis and forecast program runs. The AWN’s original list of goals was one step closer to completion.
The next step was to carry Global’s products in line-printer format back through the AWN to the High Wycombe and Fuchu forecast centers. Our present switching program required no changes to do this, and Global immediately began transmitting their long awaited products. Forecasters in the European and Asian forecast centers could now base their own prognosis charts on Global’s numerical weather prediction model.
These print-contour maps of the northern hemisphere were twelve printer pages in length – three rows of four panels each. The concept of using the AWN to deliver them was sound, but needed an important adjustment that would follow a few months later.
With Global now on-line, there still remained a number of other improvements to the system we planned to implement. One important improvement that was not completely evident at the time was in the quality of our weather intercept data. This would require some detective work a few years later and result in a dramatic increase in data quality.
A national resource
While we were still adapting NASA’s software back in DC, the Navy’s Fleet Numerical Weather Facility (FNWF) at Monterey, California, sent an emissary to visit us and he liked what he saw. Navy people at the Pentagon must have told FNWF about our new system. Soon after this visit, AWS was asked to transmit the AWN data to the FNWF via a dedicated 2,400 baud circuit the Navy would pay for. AWS agreed to do this, so Air Force footed the cost of the total system while the Navy rode piggy-back for the cost of leasing one measly circuit. Good for the Navy! I personally felt that I would be paying the Navy back for the two years of college they had given me earlier.
Again, according to Art Gulliver, “We were always aware that the Navy had an excellent [intelligence] system that kept up with what other agencies were doing in the weather business. They would get this information to their headquarters quickly and were able to exploit the advancements of others with a minimum of expense.” This perception is not meant as a criticism, rather as a compliment. We were all working together toward a common goal: the security of our country, and whenever any branch of the service could do it cheaper, the country benefited. Also, with this arrangement, we in the Air Force had the fun of working on the cutting edge of technology in our profession. The Air Force’s weather satellite program, also began in the mid-1960s, was another very successful program whose benefit was soon shared by the Navy.
The system could easily handle a new high-speed circuit without any programming changes, and we were happy to oblige. There were still plenty of unused bits in our high-speed output buffers that could be used for routing indicators. Shortly after the AWN became operational, AFCS connected the new circuit to Tinker’s communication multiplexer and we began transmitting all of our data to a grateful (and fortunate) Navy.
In late summer the National Meteorological Center (NMC) paid us a similar visit at Tinker and requested the same service. Our 418’s were still loafing over 50% of the time and we could easily feed a third high-speed circuit. With the National Weather Service as a customer, the AWN was now a national resource.
Figure 4 below depicts the AWN as it was configured in the fall of 1965.
There is more to this part of the story. NMC was in a bind. Congress had recently appropriated additional money for them, based on testimony that with this money they could implement a more sophisticated numerical weather prediction model, the Primitive Equation (PE) Model, that showed great promise of significantly improving their forecasts. Their new model eliminated certain previously necessary but debilitating assumptions about the physics of the atmosphere that prevented their present operational model to anticipate cyclogenesis (the development of new low-pressure systems – storm centers) with current acceptable accuracy. Preliminary tests of the PE model showed that it would handle this physical process well, however the PE’s much more complex mathematics would require a great deal more processing power. The only way the PE model could be run operationally was for NMC to replace their IBM 7094 system with a fifth generation computer.
At this same time, Control Data Corporation (CDC) was marketing their new CDC 6600: the fastest gun in the west. Congress gave NMC their money and a pair of 6600s were promptly leased, but unfortunately the 6600’s system software did not yet take full advantage of the 6600’s innovative hardware architecture. On-site testing with live data revealed that running the PE model on the 6600 would require at least twenty minutes more than their current model required on their 7094s. In order to begin running the PE, they would have to slip transmitting all of their products by twenty minutes. This was politically unacceptable; there would be too many complaints from influential users and Congress would probably feel duped.
Then suddenly the Air Force’s Automated Weather Network came to the rescue. When NMC linked with Tinker, Suitland began receiving sufficient upper-air data to start their analysis and forecasting runs at two hours and forty minutes past observation time rather than the previous three hours plus twenty minutes. With forty additional minutes to play with, they could now implement their PE model and still hold to their current product transmission schedule. The AWN’s value as a national resource was proportionally increased, and while this news didn’t get into the newspaper, we in the Air Force knew what happened and we were proud.
For decades, Mr. Arthur Gulliver was Global’s “Mr. Continuity” – spending most of his civil service career there. According to Art, Dr. Frederick Shuman, Director of NMC at the time, publicly acknowledged Air Force’s help. At the next joint NMC, FNWF, Global meeting he thanked AWS for their contribution and his remarks are recorded in the minutes of the meeting.
The AWN provided the Soviet Union a not-intended, but, never-the-less, significant favor. Some time after we began to send data to the National Weather Service at Suitland, Maryland, they began relaying some of our data to Europe. We somehow found out that Fuchu’s data, copied from Eastern USSR weather stations thousand of miles east of Moscow, got to Moscow before this same data which had to be relayed across the vast Soviet Union by relatively primitive means This time difference was sometimes four hours.
Phase II planning
Captain Jim Schwartz from Det. 10 would lead AFCS’s Phase II design group. He estimated that their part of Phase II would take more than a year to implement. With software and new equipment slippages it actually took two years. At our first meeting on August 12th we began scoping Phase II goals. Following is a list of the goals we sketched out:
2. Bulletin heading recognition
3. Message switching to low-speed output circuits. This meant that we would automate the manual weather relay center at all three sites.
4. Print out unrecognizable traffic at weather editor positions for correction and reentry.
5. Write all data to a magnetic tape for Project Noah’s Ark.
6. Decode all weather data types.
7. Eliminate duplicate data.
8. Build new collectives by combining shorter collectives when practicable.
9. Build all necessary weather station and bulletin heading libraries.
10. Recognize and transmit (switch) NOTAMs (FAA Notices to Airmen).
11. Assure continuity of service.
12. Facilitate through programming the supervision of high precedence traffic. Normal traffic would be transmitted on a first in-first-out basis. Hurricane and other severe weather advisories and warnings could be placed at the top of the queue.
13. Traffic analysis statistics would be maintained.
All four of our children would finally be in school that fall. This was a milestone for Almira who would now have some time to herself during the day. Doug and Joan were in Junior High, John was entering first grade, and Diane entered kindergarten. When Almira went to pick up Diane after her first day of school we were under a tornado alert. She found the children sitting against the walls in the hallways with there heads down. An unpropitious welcome to Oklahoma.
Almira volunteered as a leader of our daughter Joan’s Girl Scout troop. The following spring her troop took part in a practice civil defense emergency simulating a nuclear bomb explosion. This drill took place in a downtown Oklahoma City underground facility similar to the bomb shelters being constructed across America during that period. She related to me what transpired.
Shortly after realizing that a few of her girls were missing from the room there was a knock at the door. When she opened the door there stood her lost charges, each a budding dramatic actresses. Realistic ghastly plastic wounds were attached to them, including a rock imbedded in one girl’s forehead who screamed and fell backwards into the room. This prompted the rest of her thirteen year olds to also scream and pandemonium followed.
At another time I came home for lunch more frustrated than usual; over what I can no longer remember. As we ate, I vented my frustrations to Almira. Following a pause at the end of my tirade I suddenly smiled and added that in spite of all of my woes, this was undoubtedly the best job I ever had and I was having a ball.
Almira looked straight at me with her soft hazel eyes said poignantly “If this is true, Jack, please remember to let me know that also.”
In her non-accusing way she let me know that I had been using her as the city dump and she had been greatly concerned. If I really liked my job that much, this would make all the difference in the world to her, and it would be only fair if I remembered to remind her of this more important fact as least as frequently as I belly-ached. My wife my teacher.
Phase B – a more efficient baloney slicer
For the second time in two years Al Polston and I had been handed a major new computer software system, designed and written by someone else. For the second time in two years Al Polston and I learned the system well enough to realize that we could easily replace what was given us with a much more efficient program. This was not based on large egos, rather it was to be expected. We understood our own requirements better than authors of programs written for other purposes, and had already looked through reams of memory dumps and learned the 418’s system software well. This wasn’t too difficult, for this code contained less than 4,000 instructions – infinitely smaller than any current day vendor supplied software system. Windows 95, for instance, requires a minimum of 4 megabytes (4,000,000 bytes). Each of the 418’s instructions filled 2 and 1/4 bytes for a total of 9,000 bytes, 100 times smaller than Windows 95. Then again, the entire 418 was built with only 146,000 bytes of memory compared to my present Macintosh’s 16,000,000 bytes.
NASA’s coding was doing a terrible job separating individual weather collectives so we decided to rewrite that portion as quickly as we could. End-of-message (EOM) and start-of-message (SOM) indicators used on weather circuits, especially on Fuchu and Wycombe’s intercept circuits, were unreliable. Also, intercept circuit traffic was particularly difficult to separate because of the high garble rates on these circuits. When EOM and SOM indicators are both garbled, two different messages with two different data types will be mistakenly placed in the same message because the program could not identify their message separation point.
It was too painful for us to sit by idly while the system cried out to be improved. We could not in good conscience let this continue for two more years, so Al and I decided to proceed with an intermediate update that we dubbed Phase B. It would be a rewrite of the low-speed input logic which we would replace with what I affectionately came to call a more efficient “baloney slicer.” We sketched out an architecture much more efficient than NASA’s, but this was not surprising, for we had the advantage of having had a prototype from which we learned a great deal. Henry Ford’s first car was not a 1995 Ford Mustang! Our 1965 basic system design has lasted thirty years by now and will probably never change significantly.
I would write the main logic path, LINPRO (line processor). Over the next five years we would hang new branches (features) on this main logic trunk like ornaments on a Christmas tree while LINPRO stood firm and tall.
Each six-character input buffer was emptied – one character at a time – by a subroutine called LNMAKR that Al would write. LNMAKR’s output, a teletype line of data, would then be examined for an end-of-message (EOM) indicator or several other indications of an end-of-message. We would code for the following seven ways:
2. LLLL: an EOM signal used on FAA circuits.
3. ZCZC: the international standard SOM indicator. An SOM may not have been sent or was unrecognizable when received.
4. A bulletin heading line in case were we missed both the EOM and SOM. Bulletin headings must also be identified for switching purposes.
5. A channel sequence number: certain well disciplined circuits numbered their bulletins consecutively.
6. A circuit time-out: the transmitting station may have no further traffic for a while, and we didn’t want to hold up a completed message until a subsequent bulletin arrived.
7. Bulletin size exceeded maximum length: 200 lines overseas and 300 line at Tinker because of some lengthy gridded forecast fields from NMC.
Technical Note 2 describes with an example the challenge of writing an effective decoder. It must be neither too loose nor too tight.
Once we had checked a line of data for EOM/SOM we wrote it to our external mass storage device – a UNIVAC FH880 drum. This moved the data out of what is called the “critical line of coding” – it was no longer in our central processor’s limited memory bank where it could more easily be overwritten by data coming in too fast behind it. Unlike the computer’s memory that had space reserved for just two lines of data from any given input circuit, the drum could hold more then an hour’s worth of data.
A third task, MSGBLD (message builder), was written which would execute at a lower priority. MSGBLD would read all the lines of a given bulletin back down from the drum, verify its data type, place it in an output buffer, and hand the buffer over to AFCS for transmission.
It took less than two months to finish our goal and by mid-October we were ready to put our Phase B on-line. I would carry the new program to High Wycombe and install it and we would send the new coding on a magnetic tape to Fuchu, for Dick Wilson had the expertise to install it without our help.
During the previous three months Al and I had reassembled the whole system dozens of times and Det. 10 was beginning to be troubled over what they viewed as our too frequent updates. This problem was on my mind during my flight to England, and by the time I got to Wycombe I had thought of a clean way to separate our program sections so they could assemble and be loaded into the computer separately. It would be easy to interconnect our two parts with some simple coding. All that was required was a short table of addresses in each half that pointed to the few key points in the other program part it addressed. I would place these addresses (“transfer vectors”) in fixed locations at the top of both of our program segments. This concept worked like a charm.
I implemented this program change as soon as I got to Wycombe and called Al and gave him the changes. Al patched a phone call from me through to Dick Wilson at Fuchu on our data circuit and Dick incorporated the same changes. Suddenly we were autonomous and it felt great. This change would reduce the friction between Det. 7 and Det. 10 – at least for a while.
Due to the six hour time difference between England and Oklahoma, I worked a lot of evenings. At 10:47 p.m. London time on the evening of 9 November 1965 the high-speed circuit to Tinker was interrupted. This sometimes happened, but this interruption lasted longer than most. We later heard that the lights had gone out for 30 million people along the upper east coast of the U.S. from New York to Canada. The trans-Atlantic cable that carried our signal went ashore in the New York area and the Great Blackout of 1965 closed down the AWN for most of that night. Satellite communication technology would end this risk later on.
Writing a surface synoptic observation (SM) decoder would be our first goal. These observations are made at every weather station around the world every three hours and all of them are valuable to Global. Upper-air soundings are also valuable. Harry Rice was already intimately familiar with the synoptic code. He had written the Climatic Center’s SM decoder for ETAC’s IBM 7040 system. We asked Harry to write one for the AWN and he did his usual fine job. His decoder would run at all three sites.
Weather observations were originally transmitted in the U.S. over telegraph lines. By 1849, Professor Joseph Henry of the Smithsonian Institute established an extensive observation network after first providing telegraph companies with free weather instruments. By 1860 his network included 500 reporting station.
Telegraph line tariff rates were based on the standard of six character “groups” and the design of all weather observation codes were based on this tariff standard. The synoptic code, one of the first codes used, plus other purely numerical codes are still composed of five character words with a space character between each word which totals six characters. Today, the United Nation’s World Meteorological Organization (WMO) establishes formats for all international weather codes.
Technical Note 3 in Appendix 1 contains examples of a surface synoptic observation and a bulletin heading and shows how the syntax of each line was utilized by our decoders.
Five hole paper tape
One of the issues between AFCS and me personally was what character transmission format we would use. AFCS was required to conform to the eight-bit ASCII standard, one computer byte, but I insisted that our data stay in the form received, the traditional five-level teletype paper-tape code used on weather circuits world-wide. At that time, all weather computer systems were programmed to expect this format and I didn’t want our customers to have to convert from ASCII back to five-level. LNMAKR, for instance, couldn’t read ASCII.
Since the AWN was a dedicated weather circuit (would carry just weather data) and wouldn’t interface with any non-weather circuit, this would not create a problem so Det. 10 reluctantly agreed.
Like manual typewriters, five level teletype code uses a figure shift and letters shift characters. It also had a character for sounding a bell to alert telegraph operators, weather observers, and the like of incoming traffic.
Two carriage returns and one line feed was the standard end-of-line indicator. The second carriage return was necessary to allow time for the print-carriage to return to the left hand side of the page. When only one carriage return is received, a teletype machine will start printing its next line half way back across the page. LINPRO guaranteed two carriage returns and a line feed at the end of every data line.
I reluctantly laid my pencil down
When I returned from Wycombe, Al and I began our development of Phase II. While Al continued developing MSGBLD, I wrote a rudimentary HDSEEK; a subroutine to identify bulletin heading lines. To avoid confusion with the past, we changed the name of our mainline program from LINPRO to WXEDIT.
We were by now ready for more programming help. Reinforcements began arriving in early 1966. Ed Madigan was the first to arrive, followed by TSG Paul Mulder, and Lieutenant Jim Herod. In December 1966 my former climatology trainee colleague at ETAC Captain Ken Sleater returned from a tour in Thailand to work for me. Ken, a bachelor, made no bones about his dislike of Southeast Asia. At lunch one day he asked me “You know what I liked most about Thailand? . . . Nothing!” Ken’s sarcasm certainly didn’t convince me to volunteer.
The manual weather editing facility at Tinker, composed of a group of NCO weather observers, became a part of my unit in the spring of 1966. CMS Bobbie Hughes was the non-commissioned officer in charge (NCOIC) and MSG Roddy Smith was his assistant. These men would become valuable assets when we began to automate their function. From this group we would name librarians to maintain necessary catalogs: one with world-wide weather observing station (over 4,000) plus a weather bulletin routing library.
By the fall, our total strength hit thirty and OL-7 was upgraded to a detachment. Sometime during the middle of this growth I was forced to make a difficult decision. By now I was a major, and with all of the added administrative responsibility that comes with supervising this many people I could no longer afford the luxury of writing code alongside of Al and the others, so I reluctantly laid down my coding pencil to become a manager. I would, however, stay involved in system development planning. Even though Al was an NCO, a Senior Master Sergeant by now, I appointed him as the day-to-day technical supervisor with my full authority behind him. As usual, he did a splendid job. As soon as Captain Sleater became familiar enough with the system, he would assume Al’s role and they would develop their own working relationship. Ken respected Al’s talents and worth and never allowed it to be lost to the project.
The specific incident that precipitated this decision was when MSG Madigan asked me to make sure that paper work for his bi-annual fogy (pay-raise) was submitted on time and I realized I had no idea how to do such things. Happily, my newly assigned clerk did, but I knew then that I must spend more time taking care of the welfare of my troops. I certainly would resent my own commander if he or she left my family’s financial interests unattended.
I asked Ed Madigan to replace my preliminary HDSEEK with a genuine version. Other programmers replaced more of my preliminary code. When MSG Madigan overheard me instructing another programmer how I wanted him to improve some of my coding he commented that soon there would be no more cards in our program with my initials on them and no one in the future would know I ever worked on the project. I smiled and reminded him that I had designed the basic architecture and there had been no place to put my initials on a design, but I would always know what I had done and that would be enough satisfaction for me. He understood.
Overlays – stretching the 418 to fit the job
Much of what we were doing was brand new; had never been done before. As I have often described the experience, it was like finding money in the street. It was a true pioneering effort, and most importantly to me – it was fun!
It didn’t take us too long to realize that there was not enough space in our measly 32,000 portion of the 418’s 64k of memory, half of the 418’s central processing unit (CPU), for all the decoders to fit inside the CPU at the same time. What to do?
In a phone conversation with Global (I forget with whom – probably Dick Wilson who had been transferred to there from Fuchu by now) we heard of a concept that was new to us: overlays. It’s not necessary to keep all of the decoders in memory at once since only one is needed at any given time. Write all of our decoder subroutines to the drum and read the one you need from drum and write it on top of the decoder now in memory (overlay it). Eureka! Our mini-computer’s memory was now a balloon; we could inflate it to any size required and squeeze out all kinds of neat things heretofore unimagined. This was the final keystone required that guaranteed the ultimate success of the AWN. It took very little time for Al to write the necessary overlay management code for us.
Trips to Fuchu and Wycombe
As soon as Al finished this code we realized that it would be needed for Phase II development at both overseas sites, so we made trips to both Fuchu and Wycombe to install it. We didn’t bother to give this second mini-phase a name. In late March of 1966 Al and I flew from McClellan AFB, California, to Japan aboard a 55th Weather Reconnaissance Squadron’s mission aircraft: a Boeing WC-135 (a jet) that had long since replaced the antiquated WB-50D I had flown in six years earlier. It was like old times for me, but I had already done that, so I slept most of the way. In fact I must have slept an awful lot, for Al tells me that first we flew to Alaska and flew a Ptarmigan and from there flew a Buzzard mission to Japan. Al remembers well this, for him a once-in-a-lifetime, experience, but I don’t remember a thing – except that our first take off at McClellan took us three tries before we were airborne. Things hadn’t changed very much in six years; flying was still hazardous.
A major difference in philosophy was still outstanding between Fuchu and Det. 7. Major Telfer was convinced that the 418s would never be able to do an acceptable editing job. His concept of editing would have the 418s printout all “crippled” synoptic observations and use weather editors to manually correct any garbled weather parameter fields they found and retype entire observations back into the 418. We believed that the redundancy of intercept data made this an unnecessary waste of time and resources. I would have to have AWS resolve these differences and establish an AWN-wide standard. So it finally paid off that Det. 7 was placed directly under AWS’s commanding general.
A week after returning home from Fuchu I carried our overlay system to Wycombe by myself and installed it. Harry Rice would need this version for his almost completed synoptic decoder and Wycombe could now write and install other decoders for certain unique European codes.
The mythical man-month
During my second visit to High Wycombe I asked their forecasters what they thought of the Global products they had been receiving for nearly six months now. They said they liked the products themselves all right, but were having difficulty using them because the charts took too long getting there. Forecasters would often come on shift and have to sit twiddling their thumbs, waiting for the printer to start printing. There were also long waits between products.
I told them I would ask AWS to have Global change their mode of transmission. Currently, each product was time-scheduled at a particular time when Global thought their production system could safely guarantee delivery. This kind of schedule always has padding in it to allow for production system delays. I asked Wycombe that if Global were to change to an “as available” transmission schedule would they understand that products would still sometimes arrive late? They said they would remember and understand.
I called Lou Westphal at Headquarters as soon as I got back to Tinker and told him about Wycombe’s problem. Someone at headquarters who had the authority talked to Global who agreed to switch their transmission procedure to “as available.”
By the next week Wycombe reported that the situation had improved some, but there were still too many long delays. They also asked me why the maps printed so slowly. I explained the AWN’s lousy one-bucket brigade communication design to them. The printer spent half its time waiting a Wycombe acknowledgment to return to Tinker before Tinker could send the next 100 character message segment. (Thirty years later this one-bucket brigade sounds ridiculous.)
Their question got us thinking at Det. 7 and we began to look for a solution. I don’t remember who, but someone suggested a simple data compression technique. Presently, every one of the 132 characters on each print-line was being transmitted and this was not really necessary. There were too many instances in the data stream where long consecutive strings of the same character were sent. Most of the time it was the space character used between contours: isobars, isotherms, etc.
We agreed that trivial program changes at Global and Wycombe would solve this problem. Rather than Global insert fifteen space characters in a row, it could instead insert:
2. a phony character, a counter, to tell how many characters to reinsert, followed by
3. one copy of the character that had been removed.
In this example three characters would replace fifteen for an 83% savings. Wycombe’s print routine could just as easily be modified to check every character received to see if it was the chosen signal to refill the line. Neither change should require more than five to ten additional program instructions. They would be trivial.
We estimated that this would reduce the length of all the maps by as much as 20 to perhaps even 40% and sometimes even more, especially in warm weather when there are few intense storms and thus fewer contours (lines) on the maps. While Wycombe added water back into Global’s dehydrated maps the AWN’s one-bucket brigade would have that much more time to deliver its next segment, and the printer would begin sounding more like chunk-chunk-chunk rather than the frustrating kerchunk . . . . kerchunk . . . . kerchunk.
Lou arranged a trip for me to Global so I could coordinate these minor changes. I would modify Wycombe’s print subroutine myself and call the change to Harry Rice.
My trip to Offutt was interesting and it highlighted for me the problems that arise when managers with no programming experience are placed in charge of programming shops. It also left me with an adversary, or at least a person who would have some hard feeling toward me that I would pay for later in my career.
Global was already a large organization in 1966 and not immune from C. Northcote Parkinson’s Law and its corollaries, one of which states that “public officials tend to keep each other busy.” During my career I noticed a tendency in large offices to become overly involved with internal politics and lose touch with customers. This doesn’t come from not caring. It seems natural for managers in large organizations to drift in that direction and unless we make a conscious effort, much of our energy will be dissipated in incestuous Parkinsonian-like activities.
Unlike commercial firms, government offices don’t have salespersons to bring back complaints and recommendations from customers to better meet their needs. I believe that my visit to Wycombe served that purpose. If so, I was a good choice for the task, for by nature I am results-oriented. Most of my satisfaction at work came from seeing its ultimate benefits. That’s why I fought so hard to have the AWN keep the five-level teletype format.
At Offutt I arranged a mid-afternoon meeting with Lieutenant Colonel Butch Miller and his staff to discuss data compression. Just before going to our meeting I ambled down to the 418 computer room to take a look around. I introduced myself to the computer operator and noticed a magnetic tape hanging on the wall, and as I had already guessed, he told me it was a Global product. I asked him when he would transmit it and he said it was scheduled to be sent in forty-five minutes. As often happens, an important operational change hadn’t gotten down to the last person on shift. It was now time for our meeting so I thanked him and left.
I opened our meeting by commenting that I thought we had agreed that products would be sent as soon as they became available. Major Dick Johnston was standing, leaning against one side of a door jamb. I didn’t know who at Global was in charge of 418 computer operation. It was Dick and he answered that this had already been taken care of. Still irritated, I snapped back “Well Dick, I just left the 418 room and there was a product-tape that the operator wasn’t going to transmit for another forty-five minutes.”
Dick stood upright, barked “I’ll see to that,” and left. I’m sure he resented me for embarrassing him in front of his boss. At another time I would probably have been more considerate. Butch said nothing and we began discussing the main topic.
What transpired next highlights the problem I alluded to earlier. After I finished explaining data compression and how this technique would help our brethren at Wycombe and Fuchu, Butch nodded and turned to his assistant whom I personally knew was also not a programmer and asked him how much work such a program change would take. I was aghast at the answer – six man-months.
For a moment I sat stunned, but I couldn’t let this meeting end with such a ridiculous estimate out on the table. I also didn’t want to argue; that would only put Butch in the position of having to decide whom to believe, me or one of his own people. So I decided to try another way out of the awkward situation.
I took a chance and asked “Colonel Miller would you allow me to take a copy of your program back to Tinker and modify the code myself? I’m pretty sure I could do it in less than six months?” I didn’t hint how much less I estimated – no more than a half-hour. Butch looked at me with some amusement, shrugged his shoulders and replied “Be my guest!” He had no idea how far off that estimate was.
We ended our little tête-à-tête with a round of hand-shakes and I followed Butch’s assistant back to his office to get a program listing. As I trailed him down the hall we passed the office of some of his programmers. SSG Staff, I think his name was, came to the door and quietly whispered my name. He had been waiting for me and as I stopped and turned toward him he slipped me five program punch-cards with the changes on them. He and the other programmers somehow knew why I was there and had already coded the change . . . so much for six man-months. I quickly put the card deck in an inside pocket and continued on down the hallway. On my way out I returned to SSG Staff’s office to thank him, and he showed me exactly where in the program listing to insert the change.
I had arranged to stop at AWS the next day and was certainly glad I had done so. Colonel Roland Rogers, another extraordinarily fine senior officer, was Lou’s boss. With emotion I described to Lou Westphal and Colonel Rogers what had transpired, including being given the five card fix. Lou told me many years later that I had put Colonel Rogers in the same position of whom to believe that Butch would have been in if I had debated the estimate in his office. Global had a well deserved reputation for excellence, but the degree of excellence always boils down to the competency of a specific individual.
I arrived home Friday evening and over the weekend wrote the necessary Wycombe/Fuchu change. Monday morning I gave Offutt’s patch back to them, letting on I had written it, and arranged a test during AWN’s afternoon low-traffic test period. My patch was inserted at both Wycombe and Fuchu and Global sent one product. There was a bug in my Wycombe patch so we terminated the test. It took me just a few minutes to find the bug and I immediately called it back over to Wycombe. They corrected their program and the test map printed out perfectly on Tuesday afternoon. Having also sent the corrected code to Fuchu, we left the changes on-line and, to the delight of everyone, the shorter maps sailed through the system.
Copious thanks followed from the overseas forecasters and I gained an unwarranted reputation there and at HQ AWS as some kind of a programming genius, however, my advantage was in being an experienced programmer and any credit would more accurately be given for the hard work it took to gain this experience. So the mythical six man-months turned out to be five days.
I’ve never talked about this incident to Butch or any of his staff and no one in his office ever called me to comment either way. I have no idea whether Butch even heard the results nor whether he talked about this incident with the person who in such a convincing tone of voice advised him – “six man-months.”
Progressing toward Phase II
My job at Tinker was ideal from the beginning to end. I held the trust of my superiors and was also free of any of their pesky supervision – they were four hundred miles away. During my two years at Tinker my golf game hit a peak that I have never since attained. Because of the many hours we worked at night and on weekends I felt no guilt about taking an afternoon off for a round of golf with my people. We solved many problems with putters in hand.
Tinker was too remote for pop-in visits from headquarters. General Russell K. Pierce Jr., commander of AWS at the time, once paid us an informal visit. That evening he and his party had dinner with Almira and me at the Officers’ Club. After dancing with General Pierce, Almira repeated all the kind things General Pierce told her about the job I was doing and this felt good. I already knew that Lou and Colonel Rogers appreciated my efforts. It was gratifying to realize that they were sharing their impressions with my commanding general. Perhaps Cas’s prophesy about my career would come true.
We were allowed to shut the system down for half an hour twice a day while no upper-air data was passing through: at 2:30 a.m. and 2:30 p.m. CST. My programmers utilized most of these opportunities. Al would on occasion have to exert slight pressure to utilize this time when he felt it was required. My men were inspired, and my talents as a motivater were now being recognized. We were a formidable team.
Early one Sunday afternoon I got a call from an upset Al. As I said earlier, the 418’s belonged to AFCS so the building housing Tinker’s 418 was AFCS turf and I was about to have a skirmish with the officer-in-charge. While still in civilian clothes, Al had stopped by the building after church services that morning to check on something. The major was there and he came over and asked who Al was. Al identified himself properly, but the major told him to leave the premises immediately and never come into his building out of uniform again.
I assured Al that I would take care of the matter, and as I hung up, shook my head. Another strictly-by-the-book tin soldier. The stereotypical British comedy’s Colonel Morton J. Thwackbottom.
First thing Monday morning I called the major and he immediately began explaining that he couldn’t tell whether Al was a corporal, a sergeant, or a Navy admiral, and “this isn’t the way it should be.” I listened quietly until he finished and then, after a pause, took the offensive.
“Please, don't do this to me!” I began.
I then commenced to wax eloquently about the importance of what we were doing – revolutionizing weather communication and all that – and then I explained to him that Al had done the Air Force a favor by stopping by on his own free time to monitor this wonderful thing the major had under his care – and then I explained to him how unlikely it would have been for Al to have taken the trouble to go home and change into a uniform and return to do the same thing – and then I explained to him how often Al and my other men came into the building in the middle of the night to work (in uniform of course) – and finally, in a calmer voice, I repeated:
“Please! Don’t do this to me.”
I ended my soliloquy with a low-level but unambiguous threat . . . . If we couldn’t agree, I would have to ask my commanding general to ask his commanding general to direct that my people be allowed to stop by their work place in civilian clothing after church on Sunday only. It must have worked, for I never heard from the man again.
In October 1966 we finally had a showdown with Major Ray Telfer from Fuchu who was still adamant that Fuchu keep their current labor-intensive weather editing system. I was convinced it wasn’t needed and that our data take would not suffer if we closed it down along with Wycombe’s. We could reduce the number of editor positions to just a few who could easily concentrate on bulletin headers for data type computer editing and routing purposes.
Ray and I made our pitches to the decision makers, my bosses from AWS. Obviously I had the advantage because I worked for them and had already told them the whole story. It was only fair that Ray be given his chance to convince them, no matter how slim his chances were. Ray had been unusually insistent on this matter and wouldn’t back down nor even listen to my differing point of view.
It was no surprise that the decision was for maximum automation. Ray was furious. I can still remember how tight-jawed he was afterwards.
“ATTN DEAD HORSE THIS IS CADAVER, RGR RGR”
Al Polston has recently reminded me that we showed the good sense to utilize the experience and wisdom of our weather editors at Tinker to assist us to automate our editing section.
CMS Bobby Hughes saw a chance to train for a computer-related job after his upcoming retirement, but had tested low on the Air Force’s computer programmer aptitude test. Disappointed but not vanquished, he asked Al’s opinion and Al told him that in his experience he noticed that such results were not always a predictor of performance, that determination and commitment were more important factors. We decided to trust Bobby’s native intelligence to shine through and gave him and MSG Roddy Smith a course in ART418, UNIVAC’s 418 real-time assembly language. Both men contributed significantly to the success of the man-machine interactive techniques later used by weather editors at all three sites that enabled us to eliminate scores of boring labor-intensive blue-suit positions around the globe. Bobby Hughes was hired by UNIVAC after his retirement and later made a significant contribution to the AWN during a later upgrade from 418s to a UNIVAC 1108 computer system.
Major Bob Fanning was already at Fuchu in the weather central and replaced Dick Wilson when Dick transferred to Global. Bob and I had fun sending messages to each other over the AWN’s administrative circuit. Bob was transferred to HQ AWS headquarters after his experience in the AWN and we would coordinate AWN matters with each other until I left the system in 1970.
In 1972 Bob sent me some old yellow teletype messages we had sent to each other back in June 1967 that read:
MEMORANDUM FOR: LT. COL SHARP
JACK. CLEANING OUT DRAWERS – RAN ACROSS THIS.
THOUGHT YOU MIGHT GET A CHUCKLE.
PP EGWH RJTZ KOFF
DE WNCS 087
FM WNCS TINKER
TO ADWS HIGH WYCOMBE
UNCLAS WNCS JUNE 67. FM DET 7, HQ AWS
ATTN: LT. COL MAYKUT, MAJ FANNING, MAJ WILSON.
WE JUST RECEIVED A SPLENDID SUGGESTION FROM A/1C FRITZ, ONE OF OUR NEW WX EDITORS. HE SUGGESTS THAT WE CALL THE TINKER EDITING SECTION: “CENTRAL AEROSPACE DATA ABERRANCY VERIFICATION AND EVALUATION RELAY” OR CADAVER (C-A-D-A-V-E-R) FOR SHORT.
SIGNED, MAJ SHARP
DE RJTZ 116
152325Z JUN 67
FM 20WSD1 ADP FUCHU
TO DET 7, AWS TINKER
UNCLAS FOR MAJ SHARP. WE HAVE AN EQUALLY APPROPRIATE TITLE.
DEVELOPMENT, EVALUATION, AND DATA HANDLING (FOR) OPERATIONAL
REQUIREMENTS SATISFACTION EMITTER, OR FOR SHORT “DEAD HORSE.”
MAJOR FANNING SENDS.
UNCLAS ATTN DEAD HORSE
THIS IS CADAVER RGR RGR
An assignment to England
By January 1967 I was ripe for an overseas assignment, I had been in the states since 1958. Would I go west to Vietnam, or east to Europe, perhaps even to High Wycombe? Many of my best friends had already served a tour in Vietnam and I had mixed feelings about going there. I knew I really didn’t want to go, and could not honestly make myself believe that the whole thing was either necessary or even moral. Back in the early ‘60s there was a satirical song called “See World War Three on Paid TV,” and within a few years there we were – watching the Vietnam debacle on the evening news at supper time.
In his 1995 book, In Retrospect: The Tragedy and Lessons of Vietnam, former Secretary of Defense Robert S. McNamara’s confessed his true feelings about the hopelessness he personally felt about the Vietnam War. His confessions were shocking to me, for they reflected behavior that was the opposite to what I always viewed as loyalty to my superiors, among whom two were also McNamara’s superiors – Presidents and Commanders in Chief of the Armed Forces Kennedy and Johnson. To me, loyalty requires that I speak out, perhaps even especially, when I feel uneasy about one of my commander’s policies or decisions. Said another way, loyalty to me requires that I try to keep my boss out of trouble whenever I sense it.
As a officer in the military, I had a sense of duty that was in opposition to my true feelings about serving in Vietnam and felt some guilt, but Colonel Roland Rogers helped me out of this dilemma. I told him that I felt an obligation to my friends who had already gone to Vietnam in large numbers and I also didn’t want anyone to go to Vietnam in my place. As usual, Colonel Roger’s answer was unequivocal. “Now don’t go playing hero on me Jack, we still need you in the system.”
This helped ease my conscience a great deal for I really wanted to go to Wycombe, and happily there was another very important Air Force consideration. For a long time we knew that the High Wycombe group was being poorly managed. Phase II would be ready by the fall of 1967 and we weren’t sure that Wycombe would be ready. After all of our pressure on AFCS to rush their implementation, it would be a great embarrassment if AWS slipped implementation because Wycombe wasn’t ready. All of our prior claims of competency would now be viewed as empty boasting – braggadocio.
Lieutenant Colonel Maykut was due to return from Wycombe in June and I decided I wanted his job. When I told this to Lou he said I was crazy. Who would I work for? Why give up the title of commander so quickly, a must for a future promotion to colonel?
There were many reasons that I still can’t order in any real sense of priority. Almira’s and my decision to make the Air Force a career was based partially on the dream that my family would one day go with me to England. This vision of three years in Merry England was still alive and high on my list, but the chance to spend three years waiting for the nightmare called Vietnam to finish its grisly time upon the stage was also on the list. Another real consideration was the thought of what fun it would be to straighten out the jumble I knew existed at Wycombe. There would be fewer administrative responsibilities for me there and I could get back into the guts of the system. I was never driven by a desire for promotion, but preferred to work at what would be the most fun. If I did a good job on an important project (the AWN qualified) the promotion system would most probably reward me. If it didn’t, I would still have had the fun.
Another, perhaps bizarre but real, reason was a quirk in my personality. By spring of 1967 I felt that the cream had already been skimmed off of the AWN project; most of its (my) initial goals were realized, and the rest were designed into Phase II that was to be implemented later that year. I am forward-looking by nature and I was willing to leave the rest to others. Cleaning up Wycombe’s system would be a new challenge and more fun.
During the AWN’s early years, AWS called periodic meetings for representatives from Wycombe, Tinker, Fuchu, and Offutt, plus a Naval representative from Monterey. The next AWN system-wide meeting was convened in January 1967.
Colonel Maykut would be returning from his three year tour at Wycombe in June and would be at the meeting. He and I had agreed to try to swap jobs. By his return in June, Ed would have less than one year of service left before retiring. Ken Sleater had proven to be a valuable asset since he arrived at Det. 7 and was the obvious choice to replace me, but Maykut could be the commander until Ken took over the following year. This would also give Ken more time to learn the system.
As I left for the meeting at Scott AFB, my daughter Joan, who was the most anxious among our children to move to England, made sure that for good luck I packed my necktie with scenes of London printed on it.
During a break in our meetings, Ed walked down to the office of Colonel William S. Barney, currently AWS’s Vice Commander. Ed and the colonel knew each other from earlier times and Colonel Barney’s secretary ushered Ed into see her boss without an appointment. Fifteen minutes later Ed returned with a smile on his face; it was all taken care of. It does pay to have friends in high places.
Lou Westphal also knew how weak the Wycombe programming group was and was able to give me help in the form of CMS Jim Rosenberry. Jim who was currently assigned to the headquarters had an excellent reputation as a programmer/meteorologist and I looked forward to working with him.
Looking back, I still believe that this behind-the-scenes arranging was a great favor to AWS. At the time, no one could do what I was able to do. No one in the system except me even wanted to go to Europe for three years, and Ken Sleater was a very wise choice to be my ultimate replacement. He stayed on as commander of Det. 7 for many years and Al says that he was a fine commander in all ways and an especially competent computer system-analyst. The two of them developed a relationship similar to the one that Al and I had.
Clearing the air
Even before the Phase II was implemented in the fall of 1967 the AWN continued to grow, sometimes a little too rapidly to avoid problems. A few system hardware upgrades at both Tinker and Global proved disruptive and backlogs of traffic for Global developed at Tinker. Angers flared and accusing fingers were pointed. To clear the air, all of the parties involved had to talk to each other, so in February 1967 a meeting was called at Offutt that included HQ AFCS, HQ AWS, Det. 10, Det. 7, Global, UNIVAC, and AFCS’ Central Comm Region. Lou Westphal moderated.
A Det. 10 pilot flew us to Offutt. When we landed, the temperature was twenty below zero with a thirty knot wind from the north – it was bone-chilling cold and as I stepped off the plane the blizzard hit my face I said to myself “If I’m ever assigned to Omaha, Nebraska, I’ll swim to Cuba instead and ask for political asylum.” Three years later I would have to eat these words.
Lou had asked me to write some opening remarks for him that addressed the main issues. They were as follows:
2. We are a diverse and spread out group with differing priorities.
3. We don’t get together frequently enough. Actually, this was the first time in the twenty-month history of the system that this group had met.
4. Last December we decided to load Tinker’s input multiplexer with what we had been calling a Phase II configuration. This additional traffic coupled with routine problems while installing a lot of new electronic communication equipment caused periodic problems. There were some goofs, but each goof was corrected in a reasonable fashion, however some hard feelings remain.
5. Please discuss these incidents while you’re here.
6. Based on a joint AWS/Global decision, on January 16th all low- speed input circuits were removed from Global. This made Global totally dependent on the AWN and created problems whenever Tinker had their own problems as mentioned above.
7. There is too much traffic being sent to Tinker now to avoid backlogs as long as Offutt continues to ask for redundant data. The time has come for Global to identify their data requirements by station and data type, not by circuit, not even by bulletin heading. Self-inflicted data duplication has resulted in useful data being held up at Tinker behind data destined for no use, that is, data that has already been sent once. The AWN has the ability by now to identify data at its source 418. Global must begin using that capability or continue suffering unnecessary delays.
8. Lou mentioned also that Dick Wilson and I had spent considerable time the previous week monitoring the circuit, and agreed that there was no program problem at Tinker. Each retransmission was matched by a “nak” (non-acknowledgment) from Offutt requiring a retransmission. Dick was honest and confessed that because he had limited storage in his 418 system, he had to stop sending acknowledgments back to Tinker whenever he couldn’t forward his data to Global’s scientific computer complex.
This meeting helped clear the air of a backlog of suspicion and hard feelings. To keep the air clear, Lou and I strongly recommended that we schedule similar meetings in the future.
We were now ready for Phase II and all of its advantages that would come on line later that year after Global saw how much they were hurting themselves and finally complied with our request to respecify their requirements.
Adieu once again
I made a quick trip back over to Wycombe in April that was partly a favor to me so I could arrange for housing and take care of other personal matters. At Wycombe, Captain Vilhelm Bjerknes took me to a 17th Century farmhouse, a Quaker Hostel called Old Jordans that was close to High Wycombe Air Station. I fell in love with it, which was easy to do for it had great charm. It is located next to the Quaker meeting house where William Penn and two of his wives are buried. My wife is a Quaker but I am not, however we are both from Philadelphia, this held special meaning for the two of us.
On my return to Oklahoma I was met by a pathetic family. During my absence the five of them had gone to the base to receive all of the immunity shots required for our overseas assignment; this included a flu shot. They had all developed fevers and lay huddled en-masse on one bed for a few days, moaning in collective misery. I felt awful that I hadn’t been there to care for them. They had suffered, just as I suffered on my earlier flight to Japan.
Our daughter Joan was a quiet child, not necessarily shy, simply quiet. Inside her calm exterior there was another person waiting to be revealed to the rest of us. As we left for England, Joan told her mother of a resolve she made. When she gets to England, she explained, her new school mates will know nothing about her and won’t expect her to act the way she had always acted. Joan would be more outgoing – a different person in major ways from the child she had shown the world before leaving Oklahoma. Joan entered ninth grade in England that fall and her resolve would prove resolute.
to Cold Fronts Table of Contents
to Chapter 8