Strategic Recommendations

Robert C. Berwick, John M. Carroll, Chris Connolly, Jim Foley, Edward A. Fox, Joseph Hardin, Tomasz Imielinski, & V. S. Subrahmanian

During the latter part of the workshop, the invited PI participants (see Appendix A) enumerated a set of strategies to recommend to NSF as ways to take advantage of the opportunities presented by the Web. They then ranked the strategies by relative importance; the order in which the strategic recommendations are enumerated indicates their ranking, with the first being the most important.

Key Strategic Recommendations: We believe the most important steps for NSF to make are:

Strategic Recommendation 1: Establish NSF WWW Office

Establish and fund an NSF office to focus special attention on research issues concerning the WWW, and to help fund the implementation of selected Web enhancements which would greatly benefit the conduct of research, just as some years ago a special office was formed to handle networking. We believe the Web is of sufficient importance that this needs to be at the level of an office as opposed to a single program manager. This is the very strongest and highest priority recommendation of the workshop participants.

The most basic reason we recommend a focal-point office within NSF is the opportunity NSF has for further enhancing the impact of the Web. This impact has several dimensions. First, the impact on the conduct of general scientific research can only be enhanced, not diminished, by developing collaboration tools which work with and integrate into the Web. Second, an extensive computer science research agenda, as elaborated in Section 5, surrounds the Web. Much of this research can in some ways be done on its own, without using the Web. But, as part of the next recommendation, we argue that doing research using the Web as an extended laboratory is good science. Third, the economic development and international competitiveness aspects of increasingly sophisticated Web capabilities, driven by research with the Web, is not to be taken lightly.

Creating a focused Web office can help ensure an intentional research agenda which serves these purposes. The research agenda described in other parts of this report should fully or partially be the responsibility of this office.

Strategic Recommendation 2: Supplement PI Research in WWW Context

Target supplemental funding to PIs to do their research in the context of the WWW. For instance, a program which simulates a new chip fabrication process could be re-written to be usable via the Web. A number of the research agenda recommendations can lead to new software, such as browsers, authoring tools, and search engines - which would be of great value if widely distributed for use on the Web. A funding mechanism should be developed to support hardening of research prototypes into relatively stable (not necessarily commercial quality) distributable tools for use by a broader set of users than the immediate research community which might use a new tool on an experimental basis. In some cases, NSF should then support integration of new capabilities into non-commercial servers and browsers.

The advantages of moving research prototypes onto the Web are several experimental systems can quickly be tried out by many users. For instance, Mauldin's LycosTM search engine <URL:http://lycos.cs.cmu.edu/> receives nearly one half million requests a day as of March 7, 1995. Usage such as this can provide researchers with rapid feedback on the utility and robustness of their tools, algorithms and interfaces. Many fundamental research issues are best framed in the context of heavy usage in unpredicted ways. Understanding how real people use tools is far more revealing of their tools strengths and weaknesses than experiments in a traditional lab setting. In some sense, a Darwinian selection can occur earlier in the scientific research cycle, by exposing tools more quickly to `the light of day' offered by more general use via the Web. But, placing tools on the Web also takes more resources than working in more traditional ways.

Consider the growth in operating system research once UNIX was developed and source code was available to the research community. UNIX provided a lab environment for experimentation. The lab environment flourished even more after Berkeley UNIX was developed (with government research funding). The Internet and its preceding ARPANET had a similar effect on networking research.

A second advantage is that the availability of research prototypes on the Web increases the rate of technology diffusion by making ideas more visible, more quickly.

This is not to say that a research concept in its most embryonic stage should be tried out on the Web. Certainly the most basic ideas can and should be developed in a more controlled environment. But, researchers should be encouraged and supported in moving their embryonic systems onto the Web.

Strategic Recommendation 3: Include WWW Impact in Peer Review Criteria

Include criteria, when relevant, in peer review of proposals with respect to impact on the information infrastructure of science and engineering. Just as current review criteria include impact on science and human resources, so too, we believe, NSF should ask reviewers to consider, where appropriate, how positively the proposed research will affect the Web.

Strategic Recommendation 4: Make WWW Standard for NSF Information Dissemination

Begin an active, phased program wherein the WWW becomes the standard electronic means for NSF to disseminate information to its research communities and to the public. This is the longer-term focus of the modest beginnings discussed in our short-term recommendations which we believe NSF should make right now (and in some cases is already beginning to make). Completing the process for all of NSF, as opposed to beginning the process for a selected subset of computer-intense NSF programs or divisions, will require considerable resources and should be the subjected to debate and scrutiny.

Additional, but lower priority Strategic Recommendations: We also recommend that NSF consider whether to:

Strategic Recommendation 5: Make NSF/IRIS Projects Web Accessible

Provide funding for IRIS researchers to make current or past interactive systems accessible via the Web. This would involve retrofitting current work to the Web. It is ranked lower than the focus of Strategy 2 above because planning for Web use with new projects is potentially more efficient than retrofitting of older programs.

Strategic Recommendation 6: Develop Software Repository Tools

Develop software tools for researchers, educators, and librarians to use in creating and maintaining information and software repositories.

Strategic Recommendation 7: Establish Software Repository

Fund the establishment and operation of archives of information and software of importance to specific NSF research communities, using the software tools developed under Recommen-
dation 6.

Strategic Recommendation 8: Establish SBIR Program

Establish an SBIR (Small Business Innovative Research) program focus surrounding the Web, as a way to further jump-start economic development surrounding the Web.

Initially this should be supported by special supplemental awards. NSF should also begin to formulate revised guidelines for awards, and open up those revisions for discussion by the community to ensure enthusiastic cooperation by PIs, who will then take on these responsibilities as part of the standard obligations of awardees.

Strategic Recommendation 9: Integrate Research in Cognate Fields

NSF should encourage related research in cognate fields and integration of their results into the WWW by adding to the funding of IRIS's programs which deal with:

hypertext and hypermedia, information capture, storage and retrieval including multimedia, electronic publishing, human-computer interaction involving information access, distributed information systems, collaboration technology, knowledge models, intelligent agents, adapting to users - varied age/experience, multicultural, multilingual, computational linguistics.

This should be done in conjunction with revising program guidelines to include specific mention of the importance of applying research to the development of the WWW.

Table of Contents

Next Chapter