history of computers

Computing has totally changed how people buy and experience travel. That process seemed to start with sites that made it easy to book travel, but as with most things we experience in our modern lives, it actually began far sooner and moved down-market as generations of computing led to more consumer options for desktops, the internet, and the convergence of these technologies. Systems like SABRE did the original work to re-think travel - to take logic and rules out of the heads of booking and travel agents and put them into a digital medium. In so doing, they paved the way for future generations of technology and to this day retain a valuation of over $2 billion.
 
SABRE is short for Semi-Automated Business Research Environment. It’s used to manage over a third of global travel, to the tune of over a quarter trillion US dollars a year. It’s used by travel agencies and travel services to reserve car rentals, flights, hotel rooms, and tours. Since Sabre was released services like Amadeus and Travelport were created to give the world a Global Distribution System, or GDS. 
 
Passenger air travel began when airlines ferrying passengers cropped up in 1914 but the big companies began in the 1920s, with KLM in 1919, Finnair in 1923, Delta in 1925, American Airlines and Ryan Air in 1926,  Pan American in 1927, and the list goes on. They grew quickly and by 1926 the Air Commerce Act led to a new department in the government called Air Commerce, which evolved into the FAA, or Federal Aviation Administration in the US. And each country, given the possible dangers these aircraft posed as they got bigger and loaded with more and more fuel, also had their own such departments. The aviation industry blossomed in the roaring 20s as people traveled and found romance and vacation. At the time, most airlines were somewhat regional and people found travel agents to help them along their journey to book travel, lodgings, and often food. The travel agent naturally took over air travel much as they’d handled sea travel before. 

But there were dangers in traveling in those years between the two World Wars. Nazis rising to power in Germany, Mussolini in Italy, communist cleansings in Russia and China. Yet, a trip to the Great Pyramid of Giza could now be a week instead of months. Following World War II, there was a fracture in the world between Eastern and Western powers, or those who aligned with the former British empire and those who aligned with the former Russian empire, now known as the Soviet Union. Travel within the West exploded as those areas were usually safe and often happy to accept the US dollar. Commercial air travel boomed not just for the wealthy, but for all. People had their own phones now, and could look up a phone number in a phone book and call a travel agent. 

The travel agents then spent hours trying to build the right travel package. That meant time on the phone with hotels and time on the phone with airlines. Airlines like American head. To hire larger and larger call centers of humans to help find flights. We didn’t just read about Paris, we wanted to go. Wars had connected the world and now people wanted to visit the places they’d previously just seen in art books or read about in history books. But those call centers grew. A company like American Airlines couldn’t handle all of its ticketing needs and the story goes that the CEO was sitting beside a seller from IBM when they came up with the idea of a computerized reservation system.

And so SABRE was born in the 1950s, when American  Airlines agreed to develop a real-time computing platform. Here, we see people calling in and pressing buttons to run commands on computers. The tones weren’t that different than a punch card, really. The system worked well enough for American that they decided to sell access to other firms. The computers used were based loosely after the IBM mainframes used in the SAGE missile air defense system. Here we see the commercial impacts of the AN/FSQ-7 the US government hired IBM to build as IBM added the transistorized options to the IBM 704 mainframe in 1955. That gave IBM the interactive computing technology that evolved into the 7000 series mainframes. 

Now that IBM had the interactive technology, and a thorough study had been done to evaluate the costs and impacts of a new reservation system, American and IBM signed a contract to build the system in 1957. They went live to test reservation booking shortly thereafter. But it turns out there was a much bigger opportunity here. See, American and other airlines had paper processes to track how many people were on a flight and quickly find open seats for passengers, but it could take an hour or three to book tickets. This was fairly common before software ate the world. Everything from standing in line at the bank, booking dinner at a restaurant, reserving a rental car, booking hotel rooms, and the list goes on. 

There were a lot of manual processes in the world - people weren’t just going to punch holes in a card to program their own flight and wait for some drum storage to tell them if there was an available seat. That was the plan American initially had in 1952 with the Magnetronic Reservisor. That never worked out. American had grown to one of the largest airlines and knew the perils and costs of developing software and hardware like this. Their system cost $40 million in 1950s money to build with IBM. They also knew that as other airlines grew to accommodate more people flying around the world, that the more flights, the longer that hour or three took. So they should of course sell the solution they built to other airlines. 

Thus, parlaying the SAGE name, famous as a Cold War shield against the nuclear winter, Sabre Corporation began. It was fairly simple at first, with a pair of IBM 7090 mainframes that could take over 80,000 calls a day in 1960. Some travel agents weren’t fans of the new system, but those who embraced it found they could get more done in less time. Sabre sold reservation systems to airlines and soon expanded to become the largest data-processor in the world. Far better than the Reservisor would have been and now able to help bring the whole world into the age of jet airplane travel.

That exploded to thousands of flights an hour in the 1960s and even turned over all booking to the computer. The system got busy and over the years IBM upgraded the computers to the S/360. They also began to lease systems to travel agencies in the 1970s after Max Hopper joined the company and began the plan to open up the platform as TWA had done with their PARS system. Then they went international, opened service bureaus in other cities (given that we once had to pay for a toll charge to call a number). And by the 1980s Sabre was how the travel agents booked flights. The 1980s brought easysabjre, so people could use their own computers to book flights and by then - and through to the modern era, a little over a third of all reservations are made on Sabre.

By the mid-1980s, United had their own system called Apollo, Delta had one called Datas, and other airlines had their own as well. But SABRE could be made to be airline neutral. IBM had been involved with many American competitors, developing Deltamatic for Delta, PANAMAC for Pan Am, and other systems. But SABRE could be hooked to thee new online services for a whole new way to connect systems. One of these was CompuServe in 1980, then Prodigy’s GEnie and AOL as we turned the corner into the 1990s. Then they started a site called Travelocity in 1996 which was later sold to Expedia. 

In the meantime, they got serious competition, which eventually led to a slew of acquisitions to remain compeititve. The competition included Amadeus, Galileo International, and Worldspan on provider in the Travelport GDS. The first of them originated from United Airlines, and by 1987 was joined by Aer Lingus, Air Portugal, Alitalia, British Airways, KLM, Olympic, Sabena, and Swissair to create Galileo, which was then merged with the Apollo reservation system. The technology was acquired through a company called Videcom International, which initially started developing reservation software in 1972, shortly after the Apollo and Datas services went online. They focused on travel agents and branched out into reservation systems of all sorts in the 1980s. As other systems arose they provided an aggregation to them by connecting to Amadeus, Galileo, and Worldspan.

Amadeus was created in 1987 to be a neutral GDS after the issues with Sabre directing reservations to American Airlines. That was through a consortium of Air France, Iberia, Lufthansa, and SAS. They acquired the assets of the bankrupt System One and they eventually added other travel options including hotels, cars rentals, travel insurance, and other amenities. They went public in 1999 just before Sabre did and then were also taken private just before Sabre was. 

Worldspan was created in 1990 and the result of merging or interconnecting the systems of  Delta, Northwest Airlines, and TWA, which was then acquired by Travelport in 2007. By then, SABRE had their own programming languages. While the original Sabre languages were written in assembly, they wrote their own language on top of C and C++ called SabreTalk and later transitioned to standard REST endpoints. They also weren’t a part of American any longer. There were too many problems with manipulating how flights were displayed to benefit American Airlines and they had to make a clean cut. Especially after Congress got involved in the 1980s and outlawed that type of bias for screen placement. 

Now that they were a standalone company, Sabre went public then was taken private by private equity firms in 2007, and relisted on NASDAQ in 2014. Meanwhile, travel aggregators had figured out they could hook into the GDS systems and sell discount airfare without a percentage going to travel agents. Now that the GDS systems weren’t a part of the airlines, they were able to put downward pressure on prices. Hotwire, which used Sabre and a couple of other systems, and TripAdvisor, which booked travel through Sabre and Amadeus, were created in 2000 and Microsoft launched Expedia in 1996, which had done well enough to get spun off into its own public company by 2000. Travelocity operated inside Sabre until sold, and so the airlines put together a site of their own that they called Orbitz, which in 2001 was the biggest e-commerce site to have ever launched. And out of the bursting of the dot com bubble came online travel bookings. Kayak came in 2004

Sabre later sold Travelocity to Expedia, which uses Sabre to book travel. That allowed Sabre to focus on providing the back end travel technology. They now do over $4 billion in revenue in their industry. American Express had handled travel for decades but also added flights and hotels to their site, integrating with Sabre and Amadeus as well. 

Here, we see a classic paradigm in play. First the airlines moved their travel bookings from paper filing systems to isolated computer systems - what we’d call mainframes today. The airlines then rethink the paradigm and aggregate other information into a single system, or a system intermixed with other data. In short, they enriched the data. Then we expose those as APIs to further remove human labor and put systems on assembly lines. Sites hook into those and the GDS systems, as with many aggregators, get spun off into their own companies. The aggregated information then benefits consumers (in this case travelers) with more options and cheaper fares. This helps counteract the centralization of the market where airlines acquire other airlines but in some way also cheapen the experience. Gone are the days when a travel agent guides us through our budgets and helps us build a killer itinerary. But in a way that just makes travel much more adventurous. 

 

 

 

 

 

Qualcomm is the world's largest fabless semiconductor designer. The name Qualcomm is a mashup of  Quality and Communications and communications has been a hallmark of the company since its founding. They began in satellite communications and today most every smartphone has a Qualcomm chip. The ubiquity of communications in our devices and everyday lives has allowed them a $182 billion market cap as of the time of this writing. 

Qualcomm began with far humbler beginnings. They emerged out of a company called Linkabit in 1985. Linkabit was started by Irwin Jacobs, Leonard Kleinrock, and Andrew Viterbi - all three former graduate students at MIT. 

Viterbi moved to California to take a job with JPL in Pasadena, where he worked on satellites. He then went off to UCLA where he developed what we now call the Viterti algorithm, for encoding and decoding digital communications. Jacobs worked on a book called Principles of Communication Engineering after getting his doctorate at MIT. Jacobs then took a year of leave to work at JPL after he met Viterbi in the early 1960s and the two hit it off. By 1966, Jacobs was a professor at the University of California, San Diego. Kleinrock was at UCLA by then and the three realized they had too many consulting efforts between them, but if they consolidated the request they could pool their resources. Eventually Jacobs and Viterbi left and Kleinrock got busy working on the first ARPANET node when it was installed at UCLA.

Jerry Heller, Andrew Cohen, Klein Gilhousen, and James Dunn eventually moved into the area to work at Linkabit and by the 1970s Jacobs was back to help design telecommunications for satellites. They’d been working to refine the theories from Claude Shannon’s time at MIT and Bell Labs and were some of the top names in the industry on the work. And the space race needed a lot of this type of work. They did their work on Scientific Data Systems computers in an era before that company was acquired by Xerox. Much as Claude Shannon got started thinking of data loss as it pertains to information theory while trying to send telegraphs over barbed wire, they refined that work thinking about sending images from mars to earth. 

Others from MIT worked on other space projects as a part of missions. Many of those early employees were Viterbi’s PhD students and they were joined by Joseph Odenwalder, who took Viterbi’s decoding work and combined it with a previous dissertation out of MIT when he joined Linkabit. That got used in the Voyager space probes and put Linkabit on the map. They were hiring some of the top talent in digital communications and were able to promote not only being able to work with some of the top minds in the industry but also the fact that they were in beautiful San Diego, which appealed to many in the Boston or MIT communities during harsh winters.

As solid state electronics got cheaper and the number of transistors more densely packed into those wafers, they were able to exploit the ability to make hardware and software for military applications by packing digital signal processors that had previously taken a Sigma from SDS into smaller and smaller form factors, like the Linkabit Microprocessor, which got Viterbi’s algorithm for encoding data into a breadboard and a chip. 

The work continued with defense contractors and suppliers. They built modulation and demodulation for UHF signals for military communications. That evolved into a Command Post Modem/Processor they sold, or CPM/P for short. They made modems for the military in the 1970s, some of which remained in production until the 1990s. And as they turned their way into the 1980s, they had more than $10 million in revenue. 

The UC San Diego program grew in those years, and the Linkabit founders had more and more local talent to choose from. Linkabit developed tools to facilitate encoded communications over commercial satellites as well. They partnered with companies like IBM and developed smaller business units they were able to sell off. They also developed a tool they called VideoCipher to encode video, which HBO and others used to do what we later called scrambling on satellite signals. As we rounded the corner into the 1990s, though, they turned their attention to cellular services with TDMA (Time-Division Multiple Access), an early alternative to CDMA.

Along the way, Linkabit got acquired by a company called MACOM in 1980 for $25 million. The founders liked that the acquirer was a fellow PhD from MIT and Linkabit stayed separate but grew quickly with the products they were introducing. As with most acquisitions, the culture changed and by 1985 the founders were gone. The VideoCipher and other units were sold off, spun off, or people just left and started new companies. Information theory was decades old at this point, plenty of academic papers had been published, and everyone who understood the industry knew that digital telecommunications was about to explode; a perfect storm for defections.

Qualcomm
Over the course of the next few years over two dozen companies were born as the alumni left and by 2003, 76 companies were founded by Linkabit alumni, including four who went public. One of the companies that emerged included the Linkabit founders Irwin Jacobs and Andrew Viterbi, Begun in 1985, Qualcomm is also based in San Diego. The founders had put information theory into practice at Linkabit and seen that the managers who were great at finance just weren’t inspiring to scientists. 

Qualcomm began with consulting and research, but this time looked for products to take to market. They merged with a company called Omninet and the two released the OmniTRACS satellite communication system for trucking and logistical companies. They landed Schneider National and a few other large customers and grew to over 600 employees in those first five years. It remained a Qualcomm subsidiary until recently. Even with tens of millions in revenue, they operated at a loss while researching what they knew would be the next big thing. 

Code-Division Multiple Acces, or CDMA, is a technology that allows for sending information over multiple channels so users can share not just a single frequency of the radio band, but multiple frequencies without a lot of interference. The original research began all the way back in the 1930s when Dmitry Ageyev in the Soviet Union researched the theory of code division of signals at Leningrad Electrotechnical Institute of Communications. That work and was furthered during World War II by German researchers like Karl Küpfmüller and Americans like Claude Shannon, who focused more on the information theory of communication channels. 

People like Lee Yuk-wing then took the cybernetics work from pioneers like Norbert Weiner and helped connect those with others like Qualcomm’s Jacobs, a student of Yuk-wing’s when he was a professor at MIT. They were already working on CDMA jamming in the early 1950s at MIT’s Lincoln Lab. Another Russian named Leonid Kupriyanovich put the concept of CMDA into practice in the later 1950s so the Soviets could track people using a service they called Altai. That made it perfect for  perfect for tracking trucks and within a few years was released in 1965 as a pre-cellular radiotelephone network that got bridged to standard phone lines.

The Linkabit and then Qualcomm engineers had worked closely with satellite engineers at JPL then Hughes and other defense then commercial contractors. They’d come in contact with work and built their own intellectual property for decades. Bell was working on mobile, or cellular technologies. Ameritech Mobile Communications, or Advanced Mobile Phone System (AMPS) as they were known at the time, launched the first 1G network in 1983 and Vodaphone launched their first service in the UK in 1984. Qualcomm filed their first patent for CDMA the next year. 

That patent is one of the most cited documents in all of technology. Qualcomm worked closely with the Federal Communications Commission (FCC) in the US and with industry consortiums, such as the CTIA, or Cellular Telephone Industries Association. Meanwhile Ericsson promoted the TDMA standard as they claimed it was more standard; however, Qualcomm worked on additional patents and got to the point that they licensed their technology to early cell phone providers like Ameritech, who was one of the first to switch from the TDMA standard Ericsson promoted to CDMA. Other carriers switched to CDMA as well, which gave them data to prove their technology worked.

The OmniTRACS service helped with revenue, but they needed more. So they filed for an initial public offering in 1991 and raised over $500 billion in funding between then and 1995 when they sold another round of shares. By then, they had done the work to get CDMA encoding on a chip and it was time to go to the mass market. They made double what they raised back in just the first two years, reaching over $800 million in revenue in 1996. 

Qualcomm and Cell Phones
One of the reasons Qualcomm was able to raise so much money in two substantial rounds of public funding is that the test demonstrations were going so well. They deployed CDMA in San Diego, New York, Honk Kong, Los Angeles, and within just a few years had over a dozen carriers running substantial tests. The CTIA supported CDMA as a standard in 1993 and by 1995 they went from tests to commercial networks. 

The standard grew in adoption from there. South Korea standardized on CDMA between 1993 to 116. The CDMA standard was embraced by Primeco in 1995, who used the 1900 MHz PCS band. This was a joint venture between a number of vendors including two former regional AT&T spin-offs from before the breakup of AT&T and represented interests from Cox Communications, Sprint, and turned out to be a large undertaking. It was also the largest cellular launch with services going live in 19 cities and the first phones were from a joint venture between Qualcomm and Sony. Most of PrimeCo’s assets were later merged with AirTouch Cellular and the Bell Atlantic Mobile to form what we now know as Verizon Wireless. 

Along the way, there were a few barriers to mass proliferation of the Qualcomm CDMA standards. One is that they made phones. The Qualcomm Q cost them a lot to manufacture and it was a market with a lot of competition who had cheaper manufacturing ecosystems. So Qualcomm sold the manufacturing business to Kyocera, who continued to license Qualcomm chips. Now they could shift all of their focus on encoding bits of data to be carried over multiple radio channels to do their part in paving the way for 2G and 3G networks with the chips that went into most phones of the era. 

Qualcomm couldn’t have built out a mass manufacturing ecosystem to supply the world with every phone needed in the 2G and 3G era. Nor could they make the chips that went in those phones. The mid and late 1990s saw them outsource then just license their patents and know-how to other companies. A quarter of a billion 3G subscribers across over a hundred carriers in dozens of countries. They got in front of what came after CDMA and worked on multiple other standards, including OFDMA, or Orthogonal frequency-Division Multiple Access. For those they developed the Qualcomm Flarion Flash-OFDM and 3GPP 5G NR, or New Radio. And of course a boatload of other innovative technologies and chips. Thus paving the way to have made Qualcomm instrumental in 5G and beyond. 

This was really made possible by this hyper-specialization. Many of the same people who developed the encoding technology for the Voyager satellite decades prior helped pave the way for the mobile revolution. They ventured into manufacturing but as with many of the designers of technology and chips, chose to license the technology in massive cross-licensing deals. These deals are so big Apple sued Qualcomm recently for a billion in missed rebates. But there were changes happening in the technology industry that would shake up those licensing deals. 

Broadcom was growing into a behemoth. Many of their designs sent from stand-alone chips to being a small part of a SoC, or system on a chip. Suddenly, cross-licensing the ARM gave Qualcomm the ability to make full SoCs.  Snapdragon has been the moniker of the current line of SoCs since 2007. Qualcomm has an ARM Architectural License and uses the ARM instruction set to create their own CPUs. The most recent incarnation is known as Krait. They also create their own Graphics Processor (GPU) and Digital Signal Processors (DSPs) known as Adreno and Hexagon. They recently acquired Arteris' technology and engineering group, and they used Arteris' Network on Chip (NoC) technology.

Snapdragon chips can be found in the Samsung Galaxy, Vivo, Asus, and Xiaomi phones. Apple designs their own chips that are based on the ARM architecture, so in some ways compete with the Snapdragon, but still use Qualcomm modems like every other SoC. Qualcomm also bought a new patent portfolio from HP, including the Palm patents and others, so who knows what we’ll find in the next chips - maybe a chip in a stylus. 

Their slogan is "enabling the wireless industry," and they’ve certainly done that. From satellite communications that required a computer the size of a few refrigerators to battlefield communications to shipping trucks with tracking systems to cell towers, and now the full processor on a cell phone. They’ve been with us since the beginning of the mobile era and one has to wonder if the next few generations of mobile technology will involve satellites, so if Qualcomm will end up right back where they began: encoding bits of information theory into silicon.

Programming was once all about math. And life was good. Then came strings, or those icky non-numbery things. Then we had to process those strings. And much of that is looking for patterns that wouldn’t be a need with integers, or numbers. For example, a space in a string of text. Let’s say we want to print hello world to the screen in bash. That would be the echo command, followed by “Hello World!” Now let’s say we ran that without the quotes then it would simply echo out the word Hello to the screen, given that the interpreter saw the space and ended the command, or looked for the next operator or verb according to which command is being used.

Unix was started in 1969 at Bell Labs. Part of that work was The Thompson shell, the first Unix shell, which shipped in 1971. And C was written in 1972. These make up the ancestral underpinnings of the modern Linux, BSD, Android, Chrome, iPhone, and Mac operating systems.

A lot of the work the team at Bell Labs was doing was shifting from pure statistical and mathematical operations to connect phones and do R&D faster to more general computing applications. Those meant going from math to those annoying stringy things. Unix was an early operating system and that shell gave them new abilities to interact with the computer. People called files funny things. There was text in those files. And so text manipulation became a thing.

Lee McMahon developed sed in 1974, which was great for finding patterns and doing basic substitutions. Another team  at Bell Labs that included Finnish programmer Alfred Aho, Peter Weinberger, and Brian Kernighan had more advanced needs. Take their last name initials and we get awk. Awk is a programming language they developed in 1977 for data processing, or more specifically for text manipulation. Marc Rochkind had been working on a version management tool for code at Bell and that involved some text manipulation, as well as a good starting point for awk. 

It’s meant to be concise and given some input, produce the desired output. Nice, short, and efficient scripting language to help people that didn’t need to go out and learn C to do some basic tasks. AWK is a programming language with its own interpreter, so no need to compile to run AWK scripts as executable programs. 

Sed and awk are both written to be used as one0line programs, or more if needed. But building in an implicit loops and implicit variables made it simple to build short but power regular expressions. Think of awk as a pair of objects. The first is a pattern followed by an action to take in curly brackets. It can be dangerous to call if the pattern is too wide open.; especially when piping information For example,  ls -al at the root of a volume and piping that to awk $1 or some other position and then piping that into xargs to rm and a systems administrator could have a really rough day. Those $1, $2, and so-on represent the positions of words. So could be directories. 

Think about this, though. In a world before relational databases, when we were looking to query the 3rd column in a file with information separated by some delimiter, piping those positions represented a simple way to effectively join tables of information into a text file or screen output. Or to find files on a computer that match a pattern for whatever reason. 

Awk began powerful. Over time, improvements have enabled it to be used in increasingly  complicated scenarios. Especially when it comes to pattern matching with regular expressions. Various coding styles for input and output have been added as well, which can be changed depending on the need at hand. 

Awk is also important because it influenced other languages. After becoming part of the IEEE Standard 1003.1, it is now a part of the POSIX standard. And after a few years, Larry Wall came up with some improvements, and along came Perl. But the awk syntax has always been the most succinct and useable regular expression engines. Part of that is the wildcard, piping, and file redirection techniques borrowed from the original shells.

The AWK creators wrote a book called The AWK Programming Language for Addison-Wesley in 1988. Aho would go on to develop influential algorithms, write compilers, and write books (some of which were about compilers). Weinberger continued to do work at Bell before becoming the Chief Technology Officer of Hedge Fund Renaissance Technologies with former code breaker and mathematician James Simon and Robert Mercer. His face led to much love from his coworkers at Bell during the advent of digital photography and hopefully some day we’ll see it on the Google Search page, given he now works there. 

Brian Kernighan was a contributor to the early Multics then Unix work, as well as C. In fact, an important C implementation, K&R C, stands for Kernighan and Ritchie C. He coauthored The C Programming Language ands written a number of other books, most recently on the Go Programming Language. He also wrote a number of influential algorithms, as well as some other programming languages, including AMPL. His 1978 description of how to manage memory when working with those pesky strings we discussed earlier went on to give us the Hello World example we use for pretty much all introductions to programming languages today. He worked on ARPA projects at Stanford, helped with emacs, and now teaches computer science at Princeton, where he can help to shape the minds of future generations of programming languages and their creators.