On March 13, 2007, SirsiDynix announced that it had discontinued development of Horizon and would consolidate its future ILS efforts on Unicorn. This news sent a shockwave through the library community, further breaking the increasingly fragile ties of trust that bind libraries with the companies they rely on for software to automate their operations and to deliver their content and services. Libraries that had established an automation strategy based on Horizon face a major change in course.
In a phase of automation in which libraries are looking for new alternatives and more up-to-date technology, the death of this freshly re-engineered ILS warrants attention. This month's column gives some historical background information and offers a post-mortem analysis with some perspective and editorial opinion. Any views expressed are my own and do not reflect those of my place of work, which happens to be a Sirsi-Dynix Unicorn site.
It Began with a Troubled Birth
The library automation system we currently know as Horizon was born in 1991 and made its debut at the ALA Midwinter Meeting in January of that year. It was originally called Marquis. A small spin-off company of Provo, Utah-based Dynix Systems, Inc. originally conceived this new product in mid-1989. Marquis was the first automation system designed from its inception to follow the client/ server architecture that had emerged by the late 1980s as personal computers grew powerful enough to take on some of the workload in distributed systems environments. Marquis ran on the OS/2 operating system jointly developed by Microsoft and IBM, which at the time was positioned as the successor to Windows . Though designed for any type of library, Marquis found its early niche in the large corporate environment. Microsoft's corporate library was its first customer.
Marquis was born during a turbulent time in library automation history. Ameritech, one of the Baby Bell companies that emerged from the breakup of AT&T, decided that libraries were an interesting business niche and went on a buying spree of automation companies. Ameritech acquired OCLC's Local Systems Division (May 1990), Tacoma Public Library's Alice-B system (January 1991), NOTIS Systems, Inc. (Oct. 1, 1991), and Dynix Systems, Inc. (Jan. 20, 1992). Marquis became part of Dynix during this transition.
In this initial period, Dynix and NOTIS were allowed to operate as separate and independent business units, although wholly owned by Ameritech. Both companies offered very successful and well-established ILS products. Northwestern University originally developed the mainframe-based NOTIS system and spun it off as a for-profit company that marketed the product to other academic libraries. Dynix dominated the public library scene and had a significant presence among K-12 school and small academic libraries. The Dynix system was rich in features, stable, and very well-regarded by the libraries that used it.
Prior to their acquisition by Ameritech, NOTIS and Dynix were both based on computing platforms with waning favor, and each business unit was hard at work developing its next-generation system. NOTIS launched a project in June 1993 to develop a client/server system that ran on the UNIX operating system and the Borland InterBase relational database management system. Dynix, as I noted above, had Marquis waiting in the wings with a 4-year head start.
Following this period of independent coexistence, Ameritech unveiled its plan to consolidate its library automation businesses under the name Ameritech Library Services. NOTIS Horizon was scheduled for initial release in mid-1994. Just a few weeks before the first set of libraries was goingto place the software into production, Ameritech called a halt to NOTIS Horizon's development and withdrew it as a product. In an interesting maneuver, Ameritech changed the name of Dynix Marquis to Horizon and positioned the product as its single next-generation client/server system. Developers began to reshape this new Horizon to meet the expectations of the libraries that were expecting the NOTIS version. Ameritech ported the client software from OS/2 to Windows, which was necessary anyway given the divorce between IBM and Microsoft and the failure of OS/2 to gain traction. The company added course reserves and other features that are essential to academic libraries.
Horizon Grows and Matures
Despite this turbulent beginning, Horizon grew up to be a solid product and gained a large customer base. While its standing among academic libraries was always a bit rocky, it came to be accepted as a strong alternative for public libraries that were moving away from their aging Dynix Classic systems.
The number of libraries that adopted Horizon increased steadily, particularly in the years leading up to 2000 as institutions hurried to implement modern systems before the change of the millennium. Horizon gained in numbers through new sales and especially as libraries migrated from Dynix.
The strategy to offer Horizon as the migration path for Dynix libraries and to let these institutions move at their own pace worked out fairly well. Libraries tend to be loyal to their automation vendors. The majority choose to stay with their current vendors as they migrate from a legacy system to a new one.
Data from my lib-web-cats database demonstrates that Dynix libraries mostly stayed within the fold. While lib-web-cats isn't quite comprehensive, I believe it to be representative and unbiased. Of the 1,591 libraries listed in the database that previously used Dynix Classic, 907 migrated to Horizon, 238 selected Millennium, 200 chose Unicorn, 78 went to Polaris, and 73 moved to Library.Solution. A few migrated to other systems, and some have yet to announce their migration decisions, as lib-web-cats still lists 184 libraries using Dynix.
The Midlife Crisis and a New Beginning
Horizon was incrementally developed and enhanced through about 2004. At that time, Horizon 7.3 stood as the latest version. But more than 15 years had passed since the system's initial design. Horizon 7.4 and its Horizon Information Portal 4.0 Web-based OPAC proved to be a problematic release. The company (formerly known as Ameritech Library Services and then epixtech, and now known as Dynix) decided that this version of the software had been taken to its limit, and it was time to fundamentally re-engineer the product.
Work commenced in 2004 to completely rebuild Horizon using current-day technologies. Horizon 8.0 would be making a fresh start, largely independent of the existing code base of Horizon 7.x. Dynix procured the services of StarSoft Development Labs to team with its own programmers, thus taking advantage of StarSoft's expertise in Agile methodology for software development and inexpensive offshore programmers based in Russia. Dynix was in the later phases of Horizon 8.0 development when Sirsi acquired it in June 2005.
The newly merged SirsiDynix initially gave strong assurances that both Unicorn and Horizon would be supported and developed into the future. SirsiDynix executives touted Horizon 8.0 as the company's strategic next-generation ILS platform, while they continued to market Unicorn as a mature, stable, and feature-rich ILS. Indicators seemed to point to Horizon 8.0 should the company move toward a single ILS strategy. Given an investment estimated at more than $8 million in development costs, it seemed the natural course.
Horizon was on solid ground from a sales and marketing perspective. By 2007, it was installed in 1,597 libraries, roughly the same amount that ran Unicorn. Horizon had outsold Unicorn for the last 4 years.
Since the announcement of the Horizon 8.0 development effort in 2005, Dynix and its successor promoted this system as the forward path for libraries that were running either Dynix Classic or Horizon 7.x. The common wisdom was that Horizon 8.0 would attract the libraries that were running legacy systems from its competitors. Although tempered with a healthy measure of skepticism, the vision and direction of the new Horizon 8.0 development effort generally resonated with this group of librarians, and they made their plans accordingly. Those who made recent decisions to purchase Horizon did so based as much on the specifications of the new Horizon 8.0 platform as on the existing product. While moving to an unproven ILS would involve some risk, many libraries saw the need to migrate to current-day technology.
A Terminal Prognosis
While it's no surprise that SirsiDynix would ultimately consolidate toward a single flagship automation system, the choice of Unicorn over Horizon was unexpected. Horizon now joins the ranks of the legacy systems. I consider an automation system to be a legacy once the company announces that it will no longer be enhanced and developed. This contrasts with flagship automation systems, which have ongoing development and are actively marketed by a solvent company. To operate a legacy system, a library is usually obligated to continue making maintenance payments. The company will address day-to-day problems, implement bug fixes, and repair security-related issues but will not add major new features. When a system is declared end-of-life, the company offers no support options.
SirsiDynix states emphatically that it will continue to offer support options for Horizon 7.x for an extended period. This is not an end-of-life announcement. Horizon 8.0, on the other hand, has been withdrawn as a product. None of the beta-test libraries had yet switched to Horizon 8.0 for their production systems, and SirsiDynix won't offer it to other libraries.
This set of events in effect eliminates two ILS products: the Horizon 7.x system, which has been evolving since 1989, and the new Horizon 8.0, which has been in development since 2005.
Dying a Slow Death
The announcement that SirsiDynix will no longer develop Horizon 8.0 spells the beginning of the end for this product. From this point forward, we can expect the number of libraries that are running Horizon to diminish over time. However, these events do not force an immediate change for the vast majority of Dynix or Horizon libraries. They will have some time to review all the available options and set a new course. Naturally, SirsiDynix hopes they will migrate to Unicorn, which the company promises to rename and enhance. We can expect SirsiDynix to offer contract incentives and to launch a Herculean marketing effort to nudge Dynix and Horizon customers down that path. As I noted above, libraries tend to remain with an incumbent vendor.
Once the automation system passes from flagship to legacy status, the libraries that are running it must begin making plans to migrate. It's only a matter of time until the product reaches its end of life, and it takes librarians a long time to make plans and to find funding for their next system. Migration can be postponed until more desirable options emerge, but it's inevitable.
Even if a company offers a no-cost migration to a new flagship ILS, this course is by no means free to the library. An institution can easily expend an equal amount of its own resources in a migration as it does in capital costs for the software. The costs that are represented by such factors as the number of hours in committee meetings to manage the decision-making process for a migration, the accumulated time involved in installing the new system, migrating data, configuring policies, training staffers to operate the system, and preparing users for a new system can rival the cost of the software. Libraries have a limited number of staff resources, so all effort put toward a system migration detracts from other initiatives in which they might otherwise choose to engage.
A Disruptive Approach
Events such as SirsiDynix's decision to consolidate to a single product cause a great deal of uncertainty and disruption for libraries and detract from the trust that's necessary for a successful relationship between a library and its automation products supplier. SirsiDynix faces a large challenge in shoring up the trust with the Dynix/Horizon half of its customer base. The prompt delivery of truly outstanding replacement products is the only salve that will heal the wound.
SirsiDynix officials provide a number of arguments to justify this decision. Had the company gone forward with Horizon 8.0, it points out that all of its customers would have faced a migration at some point. For sites running 7.x, the process of getting to Horizon 8.0 is more like a migration than an upgrade, given that the product has been re-engineered from the ground up. Had SirsiDynix decided to consolidate its efforts on Horizon 8.0, the Unicorn sites would have faced a migration sometime in the future. While the folks at Unicorn sites are breathing a sigh of relief because they won't be forced into a migration in the short term, this also diminishes the opportunities to break out of their system's legacy architecture. It's yet to be seen whether libraries that are currently running Horizon will consider a lateral move to a Unicorn-based platform sufficiently beneficial and inexpensive. Many of these libraries may choose to review a wider set of options.
In the past, the library automation industry has offered a fairly gentle approach to product consolidations that result from business mergers. When a company gains ownership of multiple flagship automation systems, each with a large customer base, it's usually in that company's business interests to maintain the development of both systems for a number of years. Then, it should gently move the customer base to one of the systems or transition both systems to a next-generation product.
It's important to note that many of the companies in the industry have run a steady course. They have not put their library customers in a position of unanticipated change but have offered a fairly steady path of automation products. I can't picture a world in which libraries no longer work with commercial vendors for automation. Even the open source route would likely involve commercial companies providing service and support.
This type of event has precedents. I noted above Ameritech's abrupt abandonment of NOTIS Horizon and the repositioning of Dynix Marquis to take its place. DRA's Taos system had a similar fate. Because Taos' development languished beyond all reasonable tolerances, Sirsi's decision to withdraw it at the time it acquired DRA was understandable.
In both of these prior cases, the library community adjusted to the hand dealt by the vendors. One of the key differences in today's climate is that interest in open source alternatives for library automation has reached peak fervor.
It may well be that some libraries choose to answer the current set of circumstances with disruptive choices of their own.