Shinji Nishikawa: the Road Towards Enterprise Software Success
How Saison Information Systems came to dominate the managed file transfer (MFT) middleware segment in Japan.
The world of enterprise software is not an easy one to break into. This is particularly true in Japan, the world’s third-largest economy and home to some of the most conservative and unforgiving corporate customers.
For over forty years, Saison Information Systems has served the IT system integration needs of numerous corporate customers operating mainly in the credit card, retail, and logistics businesses. One of the company’s unexpected success stories is the HULFT series of enterprise data management products.
Originally released in 1993, HULFT’s core managed file transfer (MFT) middleware offering grew to dominate its segment in Japan. Today it is used by over 8,600 companies in 43 countries.
We interviewed, Shinji Nishikawa, former Director of Saison Information Systems, to get the story behind HULFT. This transcript has been translated from Japanese and edited for clarity.
Part 1: The Dawn
Allow me to discuss how I started to become interested in packaged software products. In the 1980s, I was in the Technology Development Department, engaged in the creation of tools for development productivity enhancement and more.
Afterward, advised to gain experience with business systems, I was appointed to undertake the credit sales management systems for Seibu Department Stores.
For half a year, I worked non-stop, without weekends, until late at night, barely catching the last train home almost every day and completed the system. However, after it was put into operation, it repeatedly ran into problems. That left me exhausted. I began to consider that it would be better to provide people with more versatile products.
I learned a lesson from that experience —sometimes, business systems can bring great pains and limited gains!
In the late 1980s, end-user computing began to spread. Operating sections, not the information systems department, started to use computers. In the 1990s, the proliferation of the internet began, which induced a shift from protocols, which varied from manufacturer to manufacturer, to standard TCP/IP. During that period, NEODAMA (a Japanese acronym referring to network, open systems, downsizing, and multi-media) was being promoted by Otsuka Corporation and became a trend in Japan’s IT industry.
Amid these changes, we were urged to adapt to open systems. However, no one was familiar with the technologies used to build them. We purchased UNIX for research purposes and embarked on a process of trial and error. Shortly after that, the section that I led in the Technology Development Department was given instructions to create a new package product.
The concept of HULFT was inspired by the words of Mr. Itano, one of my subordinates. He correctly predicted that distributed systems would become popular in the future. Until then, multi-purpose systems were used as databases for information searches, but they were inefficient and expensive. We thought they would be downgraded to open systems.
Our initial idea envisioned creating a copy of the database in a multi-purpose machine on an open system. It was what we now know as virtualization, which has become common. We wanted to create a system that would identify the database holding the data sought and carry out automatic searches following the entry of search conditions on a database.
However, it was hard to construct a virtual database with any machine available at that time. Consequently, we lowered our expectations to create a realistic product, which was the first version of HULFT. Originally, our company dealt with a number of file transfer devices in our order management systems. We used them to create a system that could automate the file transfer process and integrate with databases.
I led a team of five. We spent six months on HULFT’s conception and another half year on its development. The first version was designed to have the simple function of a one-way data transfer from the mainframe to UNIX.
HULFT originally stood for “Host UNIX Link File Transfer.”
Several prospective names were proposed for the product, but I liked HULFT because in Japanese the sound of the first syllable sounded like haru, the word for Japanese, conveying a light feeling. I also associated it with the HAL 9000, a computer with artificial intelligence in the movie 2001: A Space Odyssey.
On the other hand, any name similar to HAL Laboratory could have failed to secure trademark registration. We decided to add “FT,” the first letters in “File”and “Transfer,” at the end. Thus, the name HULFT was coined.
Part 2: A Sign of Growth
An opportunity to upgrade HULFT, arrived shortly. There was a request from Arthur Andersen, which then provided consultancy services to Seibu Department Stores concerning their POS systems.
The initial version of HULFT only supported one-way transmission from mainframe to UNIX. Andersen requested, a function for transferring files through the TCP/IP protocol and for sending and receiving data in both ways. The design of HULFT Version 2.0 was heavily influenced by this request.
At that time, our company had no engineers familiar with the C programming language. I asked the company to assign five new employees to our section to work on the HULFT upgrade. I sent them to a six-month training course outside of the company so that they could completely learn about UNIX and programing in C.
Regrettably, Seibu Department Store stopped the development of this POS system before its completion. HULFT 2 was therefore not adopted. The first user of HULFT 2 was our parent company, Credit Saison.
If I recall correctly, in those days, few mainframe products supported TCP/IP communications. We gradually received more inquiries from companies realizing HULFT’s usefulness. We received a request from Fujitsu Chiba for a version of HULFT that was compatible with their XSP. We immediately took action on this business opportunity. Our members engaged in general-purpose machines had to commute to Chiba Prefecture for almost half a year to implement the porting .
Apart from that, we embarked on the development of the Windows NT edition in a bid to broaden sales channels. Once again, we had no engineer with skills on Windows NT. Five staff members were appointed to carry out porting from the UNIX edition.
Our approach in those days was to prompt young employees to learn new technologies from scratch to broaden the range of models that we were able to support.
We actively worked to increase the number of models or operating systems that would work on UNIX if there was any request from customers.
We improved the coding method in a bid to facilitate porting to other models. To use a past example, we finished porting to another model in two weeks.
However, sales figures remained weak. HULFT was only adopted by about twenty companies. This was partly because our only channel was direct sales and we only had three staff staff conducting sales activities.
According the Fuji Chimera Research Institute, by 2010, HULFT’s share of Japan’s domestic managed file transfer market was nearly 80%. Back then, competition was fierce and there was a considerable number of competitors, including TIS and NS Solutions.
I had started to mention that our sales team could not beat Joint Net in a competition. HULFT had more functionality, but it was also more expensive. Determined to develop it into a marketable product, Toshiaki Shibayama, our then-CEO, proposed creating a low-priced edition with limited functions, under the name HULFT Dash. Its functions were limited to those covered by Joint Net for head-to-head competition.
HULFT Dash did not sell at all. After we won an RFP, the customer would end up choosing HULFT during subsequent business talks. HULFT was developing into a product selected by customers.
Part 3: Extended Function
We received an inquiry from Yokogawa Electric, which wanted an edition of HULFT for the IBM’s MVS mainframe OS. It seemed that there was a real need for data transfer between general-purpose machines and UNIX. However, our company had no IBM mainframe and judged that it would be difficult to develop the product they were seeking. Instead, we offered to disclose the source code to Yokogawa to they could develop it into what they needed.
According to the conditions of the deal, the license for the portion developed by Yokogawa would be free of charge, but the copyright and the sales rights would belong to us, and the development would be at Yokogawa’s expense. Looking back, it was surprising that Yokogawa accepted these conditions that were so advantageous to us.
However, the Yokogawa team struggled with the development because the TCP/IP APIs for Fujitsu’s IBM’s were different. As a result, the initial product was full of bugs. We worked together on debugging and achieved expansion into the IBM mainframe market.
A group company, Yokogawa Information Systems, wished to engage in sales and became our first sales agent. Soon after, a director of Yokogawa Electric advised us to cover all general-purpose machines of domestic manufacturers if we were considering overseas sales. Thus, our efforts to develop an edition for Hitachi’s systems began.
Thankfully, Yokogawa Electric proposed an introduction to Hitachi. However, Hitachi already had a data send and receive package. It was, therefore, difficult to carry out development. Instead, we were introduced to Hitachi Information Network, a subsidiary engaged in information processing.
We disclosed the source code for the IBM edition and asked Hitachi Information Network to develop a Hitachi edition under conditions similar to those for Yokogawa Electric. Accordingly, we formed an agent contract with the company that is today known as Hitachi Systems. HULFT 3 was developed amidst competition, with a view toward enhanced functionality and performance.
With HULFT 2, two-way communications were confined to those between general-purpose machines and UNIX. We revised the two-way functions, such as Chinese character conversion, which causes problems between different models. In addition, we released a Windows edition. Thus, data transfer became possible between most open models on the market and the mainframe.
Manager’s concentrated monitoring was launched to add a significant number of functions that appealed to customers. When HULFT 3 was released, we received a request to create an OS/400 edition.
BML, a company engaged mainly in blood tests and other clinical laboratory tests, planned 200 of the OS/400 units. I was skeptical about the future needs for HULFT units that were compatible with IBM’s office computers. The then-director responsible for sales promotion, Shigeru Hatta, estimated that the development cost would be recovered if 200 units were sold. That obliged us to continue.
Since our company had no one experienced in OS/400, we once again sent three personnel to outside training. At that time, no one created software that was compatible with OS/400 by combining TCP/IP and C programming language.
The trainees only learned how to operate OS/400 and how to compile in C programming language. After that, they had to learn by themselves through trial and error and carried out the development. A machine for the development was borrowed from BML. The intended unit was created after around ten months of effort.
We believed that multi-vendor support and differentiation would be decisive for the future of the product. That is why we worked from the perspective of responding to customers’ requests after we won contracts, even though we had no engineers, creating a fait accompli. We had crossed a dangerous bridge, but were desperate to grow HULFT into a marketable product.
Part 4: A Torrent of Orders
Possibly because word of mouth among manufacturers made us better known, we received requests for development from a broad range of users.
Seiyu, a supermarket chain, wanted HULFT to connect with its 200 Fujitsu K Series units operating as store servers and Tandem Computers’ NonStop System.
Matsushita Electric Works requested HULFT to connect between the K Series, serving as a factory server and the mainframe.
The K Series did not support C programming language, and we conducted development using COBOL. COBOL had some problems in developing general-purpose programs. For example, it did not support dynamic changes in record length.
We asked Fujitsu Kansai Systems (FKS), affiliated with the manufacturer and holding technologies, for help with the development. Also, FKS held talks with Matsushita Electric Works.
Development of a Tandem edition began as Seiyu requested data transfer with the Tandem used for front-end processing. We disclosed the protocol to Seiyu, and it contracted with an outsider to develop a system solely supporting binary transfer.
Then we received a request from Tandem Computers for porting of the UNIX edition after it incorporated the functions of the UNIX system into Tandem.
However, we neither had expertise in the NonStop System nor the hardware for development purposes, because a Tandem unit was expensive at any rate. Besides, we had no prospect of marketability. So, we proposed disclosing the specifications so that Tandem Computers could develop its desired system.
We prepared ourselves for the necessity of learning technologies internally, but manufacturers such as Fujitsu and Tandem Computers established their own internal structure to support development. In particular, Tandem Computers proposed that it would work on upgrades after development and for sales. That accelerated the progress.
Around that time, we gradually realized a great demand for HULFT. Version 3 was released in 1994. Then the use of UNIX machines as servers rapidly increased. Despite the burst of the economic bubble, technological innovation progressed, and IT investments in the corporate sector continued to grow.
The market saw several computer manufacturers that were each so strong that they released their own UNIX machines. They include NEC’s EWS, Fujitsu’s DS, Hitachi’s HIUX, DEC’s Alpha, and NCR’s SVR4. Models costing tens of millions of yen per unit were not unusual, but users started to introduce these diverse computers.
Under these circumstances, such customers that already used HULFT for general-purpose machines increasingly added new UNIX machines. Naturally, they wished to connect them with HULFT. We received a large number of requests for HULFT to support new models.
We thought that customers’ requests were business opportunities. Our development team worked hard every day on porting. It attained support for new models and released a model with this feature in just two weeks after the receipt of requests. It would be inconceivable nowadays.
The quantity of the units delivered soared sharply. The development personnel burned the software onto CD-Rs on their computers and packaged them for delivery. With this workflow, we were unable to swiftly ship the orders. We, therefore, introduced machines solely for burning CD-Rs and set up a team engaged exclusively in shipping. At that time, the HULFT development team had fewer than fifteen members. We were so busy working overtime every day, but we constantly upgraded the environment and increased our number of personnel to try new things.
Part 5: Becoming the De Facto Standard
The release of Windows 95 in 1995 sparked the use of the internet. With a computer, a modem and a general phone line, anyone was able to enjoy data communication. I thought that it would accelerate data exchange between companies.
Back then, inter-business communication was based on the binary synchronous communication (BSC) protocol developed in the 1960s by IBM. Of course, it did not envision the use of the internet.
Among others, the protocols recommended by the Japanese Bankers Association and Japan Chain Stores Association were commonly used. They were also heavily used in our customers’ systems. HULFT’s functions surpassed the communication environment with these old protocols. Customers wanted to use a combination of HULFT and the internet for inter-business communication.
Believing that Japan’s inter-business communication environment would be dramatically improved if HULFT replaced the rather outdated JCA protocol, w visited the associaetion of the distribution industry. When we asked, “Could you adopt HULFT in place of the JCA protocol?” We heard an entirely unexpected reply; “We are annoyed that you are creating it on your own.” The association intended to make their own infrastructure. It also envisioned making the protocol accessible solely to the distribution industry.
That stimulated our team. We decided not to rely on the industry association and worked to develop a standard communication protocol that could be used on the internet and was open to any industrial sector. That was the principal concept behind the next version of HULFT.
Many features that are nowadays commonly available, such as firewall support and EDI-CSV conversion, were incorporated into this release. We began to improve our HULFT sales promotion activities. For the logo color, we adopted red to represent our exuberance.
We actively placed ads in Nikkei Computer and other IT magazines published by Nikkei Business Publications. We took part as an exhibitor in all IT-related trade shows, such as the Open System Expo, in a bid to disseminate the product brand of HULFT.
We also worked hard on the package or the box for shipping containing the medium and the manual. We asked Mitsuo Miyahara, a designer who worked in Italy at that time, to design a package for HULFT. From among several of his designs, all related personnel, including those in the development, sales, and marketing teams, selected one that symbolized the brand image. Its look was so original that it naturally gave all personnel concerned a strong feeling of attachment to HULFT.
The company held its first product presentation event at the Tokyo Prince Hotel. We invited partners and users to the event. Thus, HULFT 4 hit the market in 1996. On that occasion, we adopted a new logo that combined the name of HULFT with the version number.
Enterprise software won’t succeed without collaboration among the development, sales, marketing, and support teams.
This version gained reactions that were entirely different from those of prior versions. Nonstop inquiries arrived from all industries. It started to be introduced to the banking sector, to which it was difficult to sell anything. With Version 4, we finally realized that HULFT was a commercially viable product.
In 1998, six years after the release, sales reached 10,000 units and the number of customers 700, both on a cumulative total basis. The HULFT Series had finally grown into a product that we could proudly dub the industry standard. Our aspiration to provide versatile products to many users had come to fruition.
