Opinion
No. 90-1264.
October 9, 1990.
Richard J. Webber, Arent, Fox, Kintner, Plotkin Kahn, Washington, D.C., argued for appellant. With him on the brief were Matthew S. Perlman and Bruce J. Moldow.
Joan M. Bernott, Sp. Litigation Counsel, Commercial Litigation Branch, Dept. of Justice, Washington, D.C., argued for appellee. With her on the brief were Stuart M. Gerson, Asst. Atty. Gen., and David M. Cohen, Director.
Cyrus E. Phillips, IV, McGuire, Woods, Battle Boothe, Washington, D.C., argued for intervenors-appellees. E. Sanderson Hoe, McKenna, Conner Cuneo, Washington, D.C., argued for intervenors-appellees. With him on the brief were John B. Dale and Charlotte D. Young. Also on the brief was Laurence Fedak, Lockheed Missiles Space Co., Inc., Fairfax, Va., of counsel.
Appeal from the General Services Administration Board of Contract Appeals.
Before MARKEY, Circuit Judge, MILLER, Senior Circuit Judge, and MAYER, Circuit Judge.
OPINION
Data General Corporation appeals the decision of the General Services Administration Board of Contract Appeals (board) granting the protest of SMS Data Products Group, Inc. (SMS) in full, and that of Lockheed Missiles Space Company, Inc. (Lockheed), consolidated with SMS's protest, in part. SMS Data Prod. Group, Inc., GSBCA Nos. 10468-P, 10474-P, 90-2 B.C.A. (CCH) ¶ 22.799, 1990 WL 31888 (Mar. 15, 1990). We reverse.
Background
The United States Department of the Interior, United States Geological Survey (USGS or agency) issued Solicitation No. 7510 on February 10, 1989. Over the seven year life of the contract proposed, the agency intended to purchase, primarily for its Water Resources Division (WRD), in excess of $125 million of automated data processing equipment, including computational workstations and other processors, peripheral hardware, software, training, support, and maintenance. The WRD is the principal federal water-data agency and is heavily dependent on information systems to fulfill its mission. Systems procured under the instant "DIS II" solicitation are intended to complement and enhance the WRD's existing Distributed Information System (DIS I), which is used to acquire, process, and store in geographic information systems enormous quantities of hydrologic data, to prepare and publish reports, and to support administrative applications.
Because the WRD employs over 4,000 people in more than 200 office locations throughout the United States, and because both the size and needs of individual WRD offices vary widely, the agency stressed in the solicitation its desire for flexibility in configuring and using the systems to be procured. In particular, the solicitation provides in section C.1.4 that "[i]ndividual systems will be custom-configured by the Government at the time each is purchased," and "[e]ach location acquiring a system may have a unique configuration, but the family of components comprising all possible systems will be common." Components to be purchased under the contract include computational workstations, each having a graphics display, keyboard, and mouse; additional processors to serve as computational or file servers; and peripheral hardware necessary, for example, to network both individual workstations and independent systems. Each "system" includes these hardware components as well as attendant software: "the total complement of hardware and software acquired for one location."
The solicitation requires that certain hardware external to the systems — for example, data gathering and output devices located in field or office, and interactive dumb terminals (IDTs), which do not themselves process information but allow their users to access and operate components of the systems — nevertheless interface and communicate with them. Similarly, systems acquired under the solicitation must be capable of communicating interactively with one another as well as processing information and using software resident on existing WRD systems. Chief among these is the DIS I, which consists of 71 Prime computers with a common operating system, common hardware functionality, and common interconnection to the agency's GEONET network.
The contractor must provide the additional hardware and software necessary to meet these connection requirements. For example, the solicitation specifies that each workstation be equipped with at least two asynchronous communications ports to accommodate IDTs, data gathering devices, data output devices, and the like. Each workstation also must have a minimum of three synchronous communications ports to allow networking. Individual DIS II system components must be connected to each other and to local DIS I hardware, like IDTs, via a Local Area Network (LAN). Remote LANs are in turn connected by wide area networks (WANs). Finally, remote WANs are connected with each other and with remote DIS I equipment via GEONET, the USGS global communications network.
The complexity of the DIS II system, and the impossibility of precisely predicting the agency's future information processing needs, dictate that USGS retain the ability to configure individual systems as local needs and unforeseen circumstances require. For instance, in small WRD offices, workstations might be able to share information processing with a computational or file server, thereby more effectively accomplishing assigned tasks. But the same approach could prove unworkable in large WRD offices, where larger numbers of users could render the link between workstations and server a "bottleneck." In this situation, an alternative solution is to provide each workstation with the necessary data storage and/or processing capability to do the job alone. To accommodate these and other possible system configurations, the solicitation requires very broad software licenses. In particular, vendors must offer licenses allowing all software to be executed simultaneously both on all processors and by all users — unlimited use site licenses. This requirement is the focus of the case.
To be technically acceptable, an offeror's system had to satisfy the technical requirements contained in section C of the solicitation, which includes the licensing provision, as well as successfully complete a benchmark test. The solicitation makes clear that the agency intended to use the benchmark "for the purpose of cost evaluation and testing" only: this or similar language is used no fewer than seven times to describe the benchmark. Attachment H to the solicitation describes the benchmark in detail; it required vendors to assemble and demonstrate the capabilities of four computer systems, consisting of 5, 15, 30, and 60 workstations, respectively. In each system, only a single user was assigned to each workstation. Although vendors could use additional processors to function as computational or file servers, none of the specified systems included or interfaced with external hardware and none was connected to another benchmark system. Only the individual workstations within each system were networked.
Data General, SMS, and Lockheed each submitted initial proposals that met the requirements of section C and passed the benchmark. Since the three received nearly identical overall technical scores, the contracting officer informed them prior to the submission of best and final offers (BAFOs) that price would control award. SMS subsequently submitted the lowest-priced offer, Data General the second-lowest, and Lockheed by far the highest. But the contracting officer disqualified SMS from the competition because, among other reasons, she determined that SMS had altered its initial proposal in two critical respects that rendered it technically noncompliant.
First, SMS proposed to split the licensing of some of its database software between workstations and servers — a "front-end back-end" approach to licensing in which one portion of a software package is licensed only for the workstations in a system and the remainder is licensed only for the servers. The contracting officer considered this a violation of the requirement that all software be licensed for all processors. Alternatively, the contracting officer found that if SMS allowed the agency to purchase full licenses for both workstations and servers, its total price would exceed Data General's. In either case, SMS was not entitled to award.
Second, in a letter sent to the contracting officer the day following submission of its BAFO, SMS "clarified" that it intended to license its software according to the number of users per system rather than the number of processors per system, as stated in its BAFO. Unlike the latter approach, in which all users of a processor are entitled to use the software licensed for that processor simultaneously, in SMS's user-based approach only a specified number of users can execute the software regardless of the number of processors to which they have access. In response to a draft SMS licensing agreement, submitted the day before BAFOs were due, the contracting officer had specifically warned SMS that the approach was unacceptable. In her view, it did not satisfy the requirement for unlimited use site licenses because SMS had not tried to determine or offered to license the maximum possible number of users of each system. Moreover, the number of users of a real system would vary and therefore be difficult to determine; generally, there would be more than the one user per workstation SMS assumed in its pricing. So even if SMS agreed to license more users per system than the number specified for the benchmark, the contracting officer concluded it would be too difficult to determine the number of licenses to purchase.
Accordingly, the contracting officer disqualified SMS from the competition and awarded the contract to Data General, which had submitted the next-lowest priced responsive offer. Both SMS and Lockheed protested to the board. See 40 U.S.C. § 759(f) (1988). SMS argued that the contracting officer improperly rejected its BAFO on the basis of its licensing provisions, while Lockheed asserted that the proposals of both SMS and Data General were deficient (for reasons other than software licensing), thus entitling Lockheed to the contract. Alternatively, Lockheed argued that the solicitation did not clearly embody USGS's true software licensing requirements and, therefore, the agency should be required to amend the solicitation and receive new BAFOs.
The board consolidated the protests and ultimately granted SMS's in full and Lockheed's in part. Because it thought "the solicitation used the words `users' and `work stations' interchangeably," 90-2 B.C.A. ¶ 22,799, at 114,493, the board determined the solicitation did not prohibit userbased software licensing of the type offered by SMS. Id. at 114,496. The board felt the agency's contrary conclusion violated Federal Acquisition Regulation 15.610(c)(3), which requires the contracting officer to "attempt to resolve any uncertainties concerning the technical proposal and other terms and conditions of the proposal." See 48 C.F.R. § 15.610(c)(3) (1989). It also counseled USGS to reassess its software licensing needs:
We believe that [the agency] should reconsider whether it really wants to be charged the inevitable higher prices for something — unlimited simultaneous use of software packages — which, in effect, it admits in its solicitation it will rarely if ever need. If it truly believes such expansive licenses are necessary, it must revise its solicitation to require such licensing terms and obtain revised best and final offers on such a basis.
90-2 B.C.A. ¶ 22,799, at 114,497. As for the requirement that all software be executable on all processors, the board concluded that the solicitation could not prohibit the front-end back-end approach to data processing that all three offerors had used for the benchmark. Id. It also questioned
whether as a practical matter it makes sense to have or pay for a complete database package resident on the 5,185 work stations in the cost model, most of which are diskless. Wouldn't it truly be "incredibly stupid" to run [relational data base management system] back-ends on diskless work stations, as testified by Data General's own expert?
Id.
To remedy the perceived ambiguities in and shortcomings of the solicitation, as well as the agency oversights from which the board believed they resulted, the board "revise[d] [USGS's] delegation of procurement authority to require [USGS], if it intends to procure these requirements, to amend the solicitation to reflect its true software licensing requirements, conduct further discussions with the offerors, and obtain revised BAFOs in accordance with the guidance provided in this opinion." Id. at 114,498 (emphasis added). Data General appeals.
Discussion
The board's decision rests on a misinterpretation of the DIS II solicitation and a misunderstanding of its own limited procurement role. Because the correct interpretation of a solicitation is a question of law, see SMS Data Prod. Group, Inc. v. United States, 900 F.2d 1553, 1555-56 (Fed. Cir. 1990), and because the board's authority to pass on an agency's automated data processing needs is circumscribed by statute, see 40 U.S.C. § 759(e) (1988), we are free to draw our own, contrary conclusions. See 41 U.S.C. § 609(b) (1988).
The board made three related errors in interpreting the solicitation. First, driven by its own assessment of the agency's "true" data processing needs, it disregarded the plain meaning of the licensing provision. That provision unequivocally requires bidders to provide software licenses not limited to any specific number of users or particular processing configuration. Second, it erroneously equated the terms "user" and "workstation," leading it to misinterpret SMS's proposal as technically compliant. Third, it focused on the parameters of the benchmark test — deliberately simplified for the purposes of demonstration and evaluation — to the exclusion of the other technical requirements of the solicitation. This error caused the board to overlook the very raison d' etre of the DIS II procurement: allowing the agency maximum flexibility in procuring, configuring, and using the hardware and software necessary to meet its manifold information processing needs. We treat each error in turn.
A. The Licensing Provision
In section C.3.3.3, entitled "Software accessibility and licensing," the solicitation provides (emphasis added):
The users of a system generally perform a wide variety of tasks and shall perform these tasks from their own WS [workstation]. Users shall be able to execute software on any processor that is part of a system in order to optimize performance of the system.
All software procured from the Contractor for a system shall be usable by a user on any WS on the system.
All software procured from the Contractor for a system shall be usable by a user on any IDT on the system. . . .
All software acquired under this contract for a system shall be executable on any processor on the system.
The Contractor shall provide an unlimited license for software use for any software package procured. This means all users of the system are allowed to execute the software simultaneously and the software may be executed on all processors that are part of the system. Although unlimited use site licenses are required, the computer system shall be used for general purpose computing. It is unlikely that a high percentage of users shall be using any one package simultaneously.
The provision could not be clearer. It requires that software licenses allow (1) all software to be executed on all processors and (2) all users to execute the software simultaneously. Since the solicitation elsewhere specifies, and the parties agree, that the term "processors" includes both workstations and servers, SMS's front-end back-end approach to licensing certain data processing software — in which it licenses a portion of the software for the workstation only and the remainder for the server only — is technically unacceptable.
Moreover, SMS would remain ineligible for award even if its proposal were interpreted to allow full software licensing for all processors. The contracting officer found, and SMS does not contest, that the additional cost of full licensing would have made SMS's proposal more expensive than Data General's. She testified:
Either I was to assume that the per server price was to be applied for all the work stations or I was to assume that we couldn't buy that piece for the work stations. There didn't seem to be anything else for me to make a determination on. So, either way, they were non-compliant. If they refused to allow us to buy that server side for the work stations, they're technically non-compliant. If they say, "Well, in fact, you can buy it, just use the server line item price for the work stations," it puts them over the top pricewise. They're no longer the low offeror. So, whichever way I looked at it, it was a problem for SMS.
Similarly, pricing software licenses on a per user basis that assumes a number of users per system, in practice, equal to the number specified for the benchmark — one user per workstation, equal to 5, 15, 30, or 60 users, depending on the system — renders SMS non-compliant with section C.3.3.3. This is so for two reasons. First, USGS had no intention of limiting itself in operation to the number of users per system specified for the benchmark. The agency stressed that the benchmark's limitations were artificial and "for the purpose of cost evaluation and testing" only. At least seven times in the solicitation the agency disclaimed any general application of the benchmark specifications. For example, the agency prefaced its "Estimated Acquisition Table," showing the approximate hardware and software requirements of the various WRD offices, with the proviso: "Although 4 system levels are required for purposes of benchmarking and price evaluation, flexibility is required in configuration of individual systems at both the initial acquisition and subsequent upgrades. Thus, each of the locations will have a unique configuration."
Second, although theoretically the maximum possible number of users could be calculated and licensed for each system ordered, SMS made no attempt to do so. It is true that the solicitation does not prohibit user-based licenses per se — licenses authorizing software use by the maximum possible number of system users are equivalent to unlimited use site licenses — but SMS limited its licenses to the number of users per system specified for the benchmark, and failed to take account of the many system users not accessing a system at one of its physically attached workstations.
Besides, given the prices per user quoted by SMS, and assuming the agency could purchase licenses for some reasonable number of real-system users in addition to the one user per workstation figure, SMS's total evaluated price would have exceeded Data General's. Although the contracting officer did not specifically so find in this context, as she did in the context of SMS's front-end back-end licensing approach, on this record the conclusion is equally valid. There is uncontroverted testimony that, assuming only three users per workstation — fewer than the maximum possible number, according to the contracting officer — SMS's price would exceed Data General's.
To the extent SMS tacitly suggested a middle ground of initially licensing the number of users per system specified for the benchmark and later granting updated licenses when and if the number of actual users exceeded this minimum, we think the contracting officer reasonably rejected it as unworkable. In light of the size and complexity of the DIS II and the unpredictable and transitory nature of its use, especially by remote users, SMS's proposed licensing scheme fails to comply with section C.3.3.3.
B. "Users"
The licensing provision itself makes clear that "users" of the system include users of IDTs as well as users of workstations. Even though IDTs are not themselves parts of the systems procured under the contract, section C.3.3.1.6.1. requires that they "operate on a system along with the WS's." In addition to the users of local workstations and IDTs, users of remote workstations and IDTs can become users of a local system via one of the WANs or GEONET. System users also include users of the existing DIS I computers, with which all DIS II equipment must interface, as well as non-WRD employees who have their own IDTs and modems — so-called "cooperators" from state geological surveys and universities who USGS regularly allows to access its computers and who, according to a WRD employee-witness, "are often larger in number than we are in an area."
In short, the board's equating of "user" with "workstation" ignores the different and much broader uses of the term "user" elsewhere in the solicitation. It also ignores the dozens of instances in which the term "workstation" refers specifically to a computational information processor, together with a graphics display, keyboard, and mouse. Substitution of "user" for "workstation" in these instances would render the solicitation nonsensical. How, for instance, can section C.3.2.11 speak of "[s]ingle-phase, 15-amp, 120-volt operation [being] required for all WS's and equipment directly connected to WS's," if a workstation is equivalent to a user, and vice versa? Contrary to the board's interpretation, the solicitation did not use the terms "user" and "workstation" interchangeably in any context — even, as discussed in section C., for the purposes of the benchmark.
Equating the terms led the board to misinterpret both SMS's initial and its final offer. SMS initially proposed to license its software on a per processor basis: its proposed response to the solicitation's licensing requirements referred to the "No. of Processors with Licensing," and the accompanying narrative stressed, "Any software procured for a system may be executed on any processor or combination of processors within that system." 90-2 B.C.A. ¶ 22,799, at 114,492. But because the board believed SMS intended to refer to system users instead, and because it thought both the solicitation and SMS used the terms "users" and "work stations" interchangeably — "even though there are places in SMS's responses where direct substitution of one word for the other would not make semantic sense" — it interpreted SMS's proposal as licensing specific numbers of users per system, dismissing SMS' use of the word "processors" as "an obvious error." Id. The board concluded that the "vagueness" in the solicitation's use of the terms "inevitably resulted in a disconnect between SMS's offer and [the agency's] interpretations of its offer." Id. at 114,496.
To the contrary, we think the terms of the solicitation are clear and SMS intended, initially, to offer per processor software licensing. Its proposal cannot be reasonably interpreted otherwise. And though it subsequently wavered, as evidenced by an exchange with the contracting officer that the board completely and inexplicably ignored, it continued to comply with the licensing requirements in its BAFO. One day prior to the submission of BAFOs, SMS telefaxed to the contracting officer a draft licensing agreement which provided, "Use of the software within a system is not restricted to particular workstations, only to the maximum number of authorized users." The contracting officer immediately and unequivocally replied that software licenses limited to a specific number of users per system were unacceptable, and that SMS should change "users" to "processors." SMS did so. In its BAFO, submitted the following day, it proposed to license software "per work station." It explained: "Licenses are priced according to the number of workstations comprising the system. . . . Payment of a license fee entitles the number of covered workstations to unlimited use of the software product. . . ." Only when the contracting officer called SMS to confirm that server licenses could be purchased on the same terms, and SMS replied that use of the term "work station" was an error because it intended to license software per user, did SMS give the contracting officer any reason to question the technical compliance of its licensing scheme. It confirmed this "clarification" of its BAFO in a December 15, 1989 letter to the contracting officer: "The software licensing prices are based on the number of users per system, not the number of CPU's [central processing units]." By thus switching to a user-based scheme that contemplated only the number of simultaneous users specified for the benchmark, SMS rendered its proposal technically unacceptable.
C. The Benchmark
We already have noted how the board ignored or minimized solicitation language limiting the role of the benchmark to technical demonstration and cost evaluation. Thus, how the terms "user" and "workstation" are employed in the benchmark is not intended to control their meaning throughout the solicitation and in SMS's proposals. We add that, even in the context of the benchmark, the solicitation did not equate or interchange the terms "user" and "workstation." It merely provided that, for the purposes of the benchmark, "a single, unique user must be assigned to and logged into each WS." That is another matter entirely. As in the rest of the solicitation, in the benchmark the term "workstation" is generally used in its technical sense — to describe a computational processor together with a graphics display, keyboard, and mouse — making substitution of the term "user" nonsensical.
The board misinterpreted the benchmark in another way as well. It concluded that, since the benchmark allowed vendors to split the processing of required relational database management systems (RDBMSs) between workstations and one or more servers, the solicitation could not prohibit the split licensing of database software as in SMS's front-end back-end licensing scheme. 90-2 B.C.A. ¶ 22,799, at 114,491. The board found support for this non-sequitur in the testimony of Data General's expert, who averred it would be "incredibly stupid" to run RDBMS "back-ends" (i.e., the backend portion of the software) on diskless workstations. Id. at 114,497. Of course it would be. The agency had no intention of doing so. Again the board ignored the solicitation's emphasis on flexibility in configuring individual systems. It also ignored the detailed and uncontroverted testimony of WRD employee Harbaugh, who repeatedly offered reasons for adding the necessary disks (i.e., memory) to workstations and running entire database programs on them, without the aid of servers:
Q: Can you explain to me why it is that your division requires the whole [database] package, both [front-end and backend]?
A: Databases are part of many of our applications. I mentioned my specialty of ground water modeling. Well, the ground water model is simulating the earth, a big chunk of the earth. You have a grid and for each grid point you have to have a data value. We store this data in the database and you have to retrieve data from that database. So we can't accept having only a fraction of our computer power which is represented by that database server as the only computer power available to store and retrieve data in the database system. We need to distribute the server part of the database function among many processors and [there is] a good chance on our systems we would really be running that server process on all of the work stations.
* * * * *
Q: Can you think of any reasons why the government would want to use the back-end of that software product on the work stations, as opposed to the servers?
A: Certainly. There are a variety of different scenarios where that could occur. . . . [O]ne situation certainly would be if they had a desire to have a work station set up that wasn't connected to any other machine, a stand-alone work station, such as you might provide a programmer or somebody like that.
Another situation would be if you had large numbers of databases which were sizeable and which were not shared by large numbers of people. Okay. For example, a scientist working on a research project involving a large number of data observations or data points, as you call them.
In that kind of a situation, if you had any substantial number of those people doing that across a network, I think you would have some real performance problems. And clearly, if you had the opportunity and flexibility to run that application on the scientist's work station, you would most likely do it.
In the DIS II procurement, the government retained precisely the type of flexibility to which Harbaugh refers. Of particular relevance to the cited passage is the board's observation that "there is no requirement for stand-alone workstations." 90-2 B.C.A. ¶ 22,799, at 114,497. That is true, but the agency certainly reserved the right under the solicitation to order stand-alone workstations for those situations described by Harbaugh. The board's independent assessment of whether the agency would or should do so is irrelevant.
It is also illegal, at least to the extent it infected the board's analysis of both the solicitation and the technical compliance of SMS's offer. Section 759(e) of the Brooks Act specifically prohibits the board from imposing on the agency its own views of the agency's data processing needs:
The authority conferred upon the Administrator [of the General Services Administration] . . . by this section . . . shall not be so construed as to impair or interfere with the determination by agencies of their individual automatic data processing equipment requirements, including the development of specifications for and the selection of the types and configurations of equipment needed.
40 U.S.C. § 759(e) (1988). Yet this is precisely what happened here. The board did not give terms either in the solicitation or in SMS's proposal their plain and unambiguous meaning, because to do so would have obligated the agency to purchase, and the successful offeror to provide, more expansive software licenses than the board thought necessary to meet the agency's needs. It said,
We believe that [USGS] should reconsider whether it really wants to be charged the inevitable higher prices for something — unlimited simultaneous use of software packages — which, in effect, it admits in its solicitation it will rarely if ever need. * * * We reach a similar conclusion with regard to the use of front-end and back-end software. * * * SMS has made a convincing case for questioning whether as a practical matter it makes sense to have or pay for a complete database package resident on the 5,185 work stations in the cost model, most of which are diskless. Wouldn't it truly be "incredibly stupid" to run RDBMS back-ends on diskless work stations . . . ?
90-2 B.C.A. ¶ 22,799, at 114,497. Accordingly, the board ordered the agency, "if it intends to procure these requirements, to amend the solicitation to reflect its true software licensing requirements. . . ." Id. at 114,498. "Stupid" or not, the board has no warrant to question the agency's judgment or to revise its delegation of procurement authority to ensure that the agency's assessment of its "true" needs is in harmony with the board's. The board has neither the authority nor the expertise to secondguess the agency.
Conclusion
Accordingly, the decision of the board is reversed and the award to Data General is reinstated.
REVERSED.