CORNELL UNIVERSITY

MESL TECHNICAL REPORT

Institutional Profile

Organizational Profile

Cornell University

Cornell University opened in 1868 as a nonsectarian institution of higher education in fulfillment of the dream of its founder, Ezra Cornell, an adventurous businessman. Cornell's goal was to create "an institution where any person can find instruction in any study." From its start, the University has offered courses of study in both practical and theoretical subjects.

Today Ezra Cornell's vision of an egalitarian institution open to all continues to inspire education at Cornell University. The University currently includes thirteen colleges and schools. On the Ithaca campus are the seven undergraduate units--the College of Agriculture and Life Sciences; the College of Architecture, Art, and Planning; the College of Arts and Sciences; the College of Engineering; the School of Hotel Administration; the College of Human Ecology; and the School of Industrial and Labor Relations -- as well as four graduate and professional units: the Graduate School, the Law School, the Johnson Graduate School of Management, and the College of Veterinary Medicine. The Medical College and the Graduate School of Medical Sciences are in New York City. Because Cornell is an Ivy League university and also the land-grant institution of New York State, it is a unique combination of public and private divisions. Stemming from the university's land-grant role are Cornell Cooperative Extension (an education-outreach program for New York State residents) and the belief that the fruits of Cornell research should extend into the public domain.

The university's 13,510 undergraduates and 5,970 graduate and professional students come from all fifty states and more than a hundred countries. Undergraduate students can choose from among four thousand courses, and graduate study is available in about a hundred academic and professional fields.

Interdisciplinary study and research are Cornell hallmarks, as is attention to undergraduate education. The university's 2,340 faculty members are active teachers as well as researchers--Nobel laureates often teach introductory courses--and the lines of traditional disciplines are easily crossed. Engineering students dabble in photography; theater arts students explore the world of computers; physics students learn landscape architecture.

Cornell ranks eighth among United States universities in total research and development activities, and second in funds allocated by the National Science Foundation (NSF) for programs in academic science and engineering. It is the site of seven national research centers

International programs at Cornell include a remarkable cross section of area and topical studies. Five have been designated by the United States Department of Education as national resource centers. Cornell teaches more than fifty languages and was the first university to teach modern Far Eastern languages.

The Cornell University Library supports and enhances the university's teaching and research programs. The seventeen Ithaca-campus libraries provide a rich array of reference and information services. The Library holds more than 6 million volumes, 6.5 million microforms, 211,000 maps, 71,000 sound recordings, 53,000 cubic feet of manuscripts and archives, 3,000 computer files, and 1 million paintings, prints, and photographs. Additionally, the Library subscribes to 62,000 journals and other serial publications. The Cornell Library is an international leader in the development of networked access to electronic resources and has conducted pioneering studies in the application of digital imaging technologies for preservation and access.

The Herbert F. Johnson Museum of Art is a major visual arts center for upstate New York, with notable Asian, American, and graphic arts collections. It has a national reputation for its diverse range of high-quality exhibitions and serves as an important educational resource for the Cornell community.

The Digital Access Coalition

The MESL implementation at Cornell has been conducted under the auspices of the University's Digital Access Coalition. The Coalition is a cooperative effort by librarians, archivists, curators, technologists, and teaching faculty from across the campus to develop a new vision for organizing, accessing, and using the University's various historical, ethnographic, artistic, and natural history collections. The Coalition was established in 1992 with the support of the University Librarian and the Vice President for Information Technologies and is now sponsored by the University Provost.

The goals of the Coalition include:

General Description of Computing Environment

The campus data network, developed through a joint project with IBM, has evolved to include a 100 megabit core FDDI ring using dual-homed concentrators for redundancy and fault tolerance, linked to nine subsidiary FDDI rings. The backbone uses approximately 100 IBM 6611 routers in 65 buildings (servicing approximately 300 buildings total), and linking 450 LANs and 15,000 10baseT Ethernet end stations (8,000 in academic and administrative areas, including public labs; 6,000 in Residence Halls; and an additional 1,000 LocalTalk nodes which are in various stages of transition to Ethernet). Cornell became an early adopter of 10baseT, when in 1989 an RFP was issued for Ethernet over phone wire shortly before the 10baseT standards were finalized. Since extra wire had been installed along with the phone system in 1985, there was no need to redo the wire plant.

The FDDI routers use GateD routing software running OSPF routing protocols. GateD is software developed at Cornell by the Cornell GateDaemon Consortium to provide a scaleable solution for routing on large internets. The networking protocols routed on the Cornell backbone are IP and AppleTalk. An AppleTalk filtering scheme, also developed at Cornell, provides additional security.

Dial-in facilities are connected to the campus backbone, and support both SLIP and serial terminal connections. The primary connection to the Internet for most campus users is provided by NYSERNet, and employs a Cisco 7000 router connected to a NYNEX T3 circuit, which in turn connects to Sprint's facilities in Syracuse, New York. Cornell also has a T1 link to the Geneva Experimental Station in Geneva, New York.

The MESL implementation uses servers in the Center for Theory and Simulation in Science and Engineering (also known as the Cornell Theory Center). Image data is stored on disk drives, and the MESL data resides on a Macintosh server.

MESL was implemented as a World Wide Web site. Students had access from anywhere on campus, or via dial-in, including classrooms, computer laboratories run by Cornell Information Technologies, and a gallery in the Museum. Most small classrooms in the Arts Quad do not have IP connections to the campus network. As a result, use of the MESL images in classrooms was limited to large lecture halls which did have IP connections and into which computers, monitors, and projection equipment could be brought (there are no classrooms in the Arts Quad with built-in computer projection equipment). There were no problems associated with the use of portable computer and projection equipment. More than one faculty member had a monitor that could not support millions of colors, limiting their use of the MESL images from their desks.

User Demographics

There are 13,510 undergraduates, 5,970 graduate and professional students, and 2,340 faculty members at Cornell University. All with access to the Cornell IP domain had access to the MESL images.

STAFFING

Organization of MESL Project Team

List roles and responsibilities with respect to MESL, number and level of staff devoted to MESL technical processing, changes, year 1 to year 2, and approximate amount of time spent by each staff person listed.

Comments:

The staff identified above did the primary work associated with Cornell's MESL implementation. The remainder of the MESL team is identified in the following item.

Others in organization that "did the work" of MESL

List roles and responsibilities with respect to MESL, number and level of staff devoted to MESL technical processing, changes, year 1 to year 2, and approximate amount of time spent by each staff person listed

Comments:

A large number of faculty, staff, and organizational units contributed to the success of the MESL project at Cornell. This report assesses the technical effort expended in implementing and MESL at Cornell. Cornell faculty and staff also participated in the MESL working groups, contributed to the development of the MESL project as a whole, and supported a wide range of evaluation efforts. Cornell participants in the MESL effort include:

College of Architecture, Art & Planning

Margaret Webster, Director of the Visual Resources Facility

Jean Locey, Chair of the Art Department & Professor

Sheryl Reiss, Lecturer/Research Associate, Art Department

College of Arts & Sciences

Claudia Lazzaro, Chair of the History of Art Department & Professor

Laura Meixner, Acting Chair of the History of Art Department & Professor

Sheryl Reiss, Lecturer/Research Associate, History of Art Department

Cornell Theory Center

Marcy Rosenkrantz, Associate Director, Supercomputing Technologies

Laurie McEvoy, Systems Programmer

Cornell University Library System

Tom Hickerson, Director, Division of Rare & Manuscript Collections (RMC)

Elaine Engst, University Archivist, RMC

Noni Korf Vidal, Digital Projects Archivist, RMC

Barbara Prior, Reference Librarian, Fine Arts Library

Digital Access Coalition

Tom Hickerson, Director

Peter B. Hirtle, Assistant Director

Interactive Multimedia Group (IMG)

Geri Gay, Director, and Professor, Department of Communication

Robert Rieger, Coordinator

Noni Korf Vidal, Designer/Researcher

Herbert F. Johnson Museum of Art

Carole Denatale, Registrar

Training needs

Did you encounter any specific training needs required to perform MESL technical tasks?

The technical staff needed to learn a new set of software to manipulate and publish the images. Software that had to be mastered included DeBabelizer, Tango, and CD-R mastering software. In addition, staff had to learn HTML so that they could write out HTML code from Excel, and also use several different HTML editors.

Describe how those needs were met.

Staff were all self-taught, learning from the tutorials and manuals accompanying the software.

COMMENTS

In spite of efforts to deliver a "standard" image and database product, staff discovered that the MESL data and images required extensive manipulation before they could be "published" on our web site. This meant staff had to become proficient with tools that could make the batch manipulation of data and images relatively easy.

Technical Implementation

Description of the MESL Delivery System

This section is intended to produce an outline of the technical implementation of each campus' MESL delivery system and procedures.

MESL Integration into Information Services

Describe how your MESL delivery system fit with other campus services. Was it just another offering on an existing menu of information services (and if so, generally describe the set of services, who manages them, and how MESL differs from the others)?

The MESL implementation at Cornell exists independently from other information services. The underlying search engine is separate from the Library and other information services on campus. Users were directed to the MESL service through links on related web pages. The Fine Arts Library and the Rare and Manuscript Collections both included links to MESL from their general Web home pages. There was also a link from the page for Utopia, an in-house image database. And of course MESL was indexed by the general Cornell Web index software.

Many of Cornell's electronic resources are cataloged in the Cornell Library catalog. There was some talk of including a record for MESL, but this was rejected for two reasons. First, the MESL images were sure to be available on campus only for a relatively brief time. Secondly, Cornell has only implemented a test web client to its catalog, making it difficult to link directly from a catalog record to a web resource such as MESL. The lack of Z39.50 functionality in the Cornell MESL implementation limited how much it could be integrated with traditional library information services.

Does your institution view the MESL images as part of a larger body of information resources?

Yes and no. We do think of the MESL as part of the body of information resources available to the students, but because MESL was only a two-year project, we have limited somewhat our efforts to integrate the MESL images into the larger body of information resources.

System Architecture

Server

Hardware used for MESL server, storage

The database and thumbnail images are on a Quadra 700 with 20 megs of RAM and a 1 GB hard drive. The screen and large images are on a IBM RS6000 C10 Power PC. Both of these machines are housed at the Cornell Theory Center.

Operating System

MacOS 7.5.3

Server software

Webstar

Server Performance

Did the "implementation" of MESL result in noticeable degradation of server performance? Was MESL part of a rationalization to upgrade the server (or install a new server) ?

MESL was the only application on this server. The server is slow, which did serve as a justification for the purchase of a new Windows NT server. Assuming that MESL continues past the deadline, the MESL data will probably be moved to the new server.

Client types

Web browsers supported

Any HTML 3.2-compliant browser.

Other clients

The initial distribution of the MESL data was on CD-ROMs with Shoebox software loaded. No effort was made to replicate this arrangement after the second distribution of data and images. The MESL application became purely web-based, and only web-browsers can be used to access the database.

Storage

Amount of disk space devoted to MESL data

Storage space for the MESL project can be divided into several different classes of data:

MESL data available only off line (e.g. on distribution CDs)

The supporting text documents distributed with the MESL images have never been loaded into the system, and are only available on the distribution disks. Also, the high resolution images on the PhotoCD distributed by the Eastman House were not made available; they are only available on the original disks.

Incremental Storage Requirements

Did you need to buy new storage for the MESL project? Initially? In the course of the project? If so, approximately how much did it cost?

Two 4 GB drives were purchased for the project. Each drive cost approximately $1,300, for a total of $2,600.

Data Preparation and Loading

General

Was the MESL data incorporated into an existing database application or was it treated as a separate (new?) application?

A new application was developed for MESL data.

Image Preparation

Did you convert images to other sizes, formats, and resolutions?

Yes.

If you did convert images to other sizes, formats, etc., please describe the sizes and formats you chose, and indicate for what purpose the images were converted.

The images were converted to make it possible to display thumbnail images with the data, to make it easier to browse, and to assist in the rapid identification of images. For thumbnails, the longest dimension was scaled to 120 pixels. Medium-resolution images with 390 pixels devoted to the longest dimension were also created. Both of these images were saved as JPEG images. The highest resolution images were left either as a TIFF or JPEG image, or converted to JPEG from PhotoCD, depending upon the source format.

If you compressed images as part of the conversion process, what quality setting/compression level did you use?

High quality/low compression

Image Preparation Software

List all the software your project team used to convert the images (including changing sizes and compressing). Note whether you bought the software during the course of the project or just used something already on-hand. For each piece of software note the specific functionalities you used and whether you used this software a little, periodically, or a lot. Give the approximate price of the software.

Comments:

If you could start over today, would you use this software or something else?

We would continue to use this software.

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of skill-sets they needed to effectively accomplish this. Estimate the amount of time each person spent on this phase for each distribution, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

Structured Data Preparation

What were the stages you went through to prepare the structured data for mounting on your campus network? Who was involved? when? What software did you use to process the text descriptions? Please describe how you did this.

The MESL data were first imported into an Excel spread sheet. This gave the MESL project coordinator (who did the data conversion and clean-up) an overview of the data and problems it might present. The data were then cleaned in Excel. Problems with poorly (or undelimited) fields, character sets, and other anomalies (such as delimited data with line feeds in the middle of the record) were identified. The fact that different institutions used different delimiters greatly complicated this portion of the project. Once cleaned, the data were imported into FileMaker Pro software for web access. In addition the initial distribution of images and data was also imported into Kodak's Shoebox software.

Because the first distribution of data were redistributed, staff had to go through this clean-up process twice the first year.

Structured Data Preparation Software

List all the software your project team used to convert the structured text (including parsing, recoding, and renaming fields). Note whether you bought the software during the course of the project or just used something already on-hand. For each piece of software note the specific functionalities you used and whether you used this software a little, periodically, or a lot. Give the approximate price of the software.

Comments:

If you could start over today, would you use this software or something else?

The software has been acceptable. We might implement more of the relational features of FileMaker Pro 3.0. We followed instead a flat-file model as implied by the MESL data dictionary.

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of skill-sets they needed to effectively accomplish this. Estimate the amount of time each person spent on this phase for each distribution, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

One of the great disappointments of the MESL project was the amount of time required for cleaning and restructuring data in order to be able to import it into the database and deliver it to the campus.

Unstructured Text Preparation

Did you make unstructured texts available? How were they linked to the structured records? Did you perform any additional processing to make them available?

We did not make the unstructured text available.

Describe the process of linking the unstructured text to structured data, including difficulties encountered.

N/A

Unstructured Data Preparation

List any problems you had (such as loading and handling of filenames and storage arrangement) with preparing the unstructured data.

N/A

If you provided access to unstructured data via methods other than simple links, please describe.

N/A

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of skill-sets they needed to effectively accomplish this. Estimate the amount of time each person spent on this phase for each distribution, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

Functionality of MESL System

Browsing

Did you provide browsing capability in your MESL implementation? Describe.

Yes. Faculty and students were encouraged to create subsets of the MESL data for particular courses or projects. A "browse" of one of the subsets or of images from an institution retrieves a page with ten to fifty thumbnail images on the page.

Searching

Describe the search functionality provided in your MESL implementation.

Search engine

The search engine is the one built into FileMaker Pro.

Fields indexed

Holding Institution

Creator Name

Creator Culture

Creator Role

Creation Place

Creation Technique

Associated Events, People, Organizations & Places

Concepts-Subject

Concepts-Function

Concepts-Style/Period

Object Title

Object Type

Begin Date

End Date

Material Medium

Image Filename

Display options for search results

Either a thumbnail image with brief text or only text can be selected. The number of records and/or images to be displayed can be modified, or the result set can be sorted.

Modifications to Data Dictionary

Did you change fieldnames from the MESL data dictionary (and if so, why)?

Yes, the names of some fields were changed. It was felt that the default MESL field names might confuse potential users. An effort was made to translate the terms into labels Cornell users would understand.

Types of searching supported (Boolean, content-based, etc.)

Both fielded-text and keyword searching are supported. The only Boolean operator permitted is "and." There is default truncation in searching, but a search for an exact match is possible. Date range searching is also possible.

Software

What software did you use for these functions? Did you use off-the-shelf applications or build things from scratch yourself?

The underlying database is FileMaker Pro 3.0. It is accessed from the web using ROFM CGI shareware.

If you could start over today, would you use this software or something else? Please explain

We are considering moving the application to Microsoft Access database running on a Windows NT machine. If we were going to keep it on a Macintosh platform, we would use Tango as the interface to the Web.

Other functionality

Did you add tools to provide additional functionality? (such as slide-sorting, zoom and overlay functions, templates to aid in building applications, etc.). Were these tool developed in-house or derived from off-the-shelf applications?

Professors and students were provided with templates for creating image review sites and/or galleries.

What kind of additional tools would help the instructor or student's use of MESL type data

These are the kind of things we would like to add to our MESL implementation:

1. A "light table" approach to browsing, that permits the display of multiple thumbnails during browsing (8 at a minimum), and allows one to select from those thumbnails to build a new search set.

2. The ability to display all the terms indexed for any particular field from within the search environment. The lack of a shared indexing vocabulary made searching the MESL database difficult. Faculty therefore asked for the ability to browse the terms that were used. We provided this within the site but was not integrated in an interactive manner.

3. Images that can be interactively resized to accommodate the width of the screen.

4. The ability to zoom in on images.

5. Extreme rapidity in image retrieval and display. Faculty and students want to be able to browse large number of images very quickly.

6 The ability to select two images and display them side by side for comparison.

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of skill-sets they needed to effectively accomplish these. Estimate the amount of time each person spent on this phase, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

System Security and Authentication

General Security

Describe how you provide security for MESL data to conform to the terms of the cooperative agreement.

The MESL server is restricted to IP addresses from the Cornell domain. Outsiders can have access to the general information pages about MESL, but it requires a Cornell IP address to actually retrieve MESL images and/or data.

Authentication

Did you implement a specific authentication scheme? If so, describe.

No.

Security

Were your security steps taken specifically for MESL? Did you previously have this type of security? Did MESL provide the rationale for undertaking security measures you think you'll need to undertake in the future anyway?

Because the server used was dedicated to the MESL project, the IP restriction on the server was implemented specifically for MESL. This type of security had been used at Cornell in the past. Future security implementations may be based on the security systems used at Cornell (primarily Kerberos).

Software

What software did you use for these functions? Did you use off-the-shelf applications or build things from scratch yourself?

Webstar server software offers the ability to restrict by IP address.

If you could start over today, would you use this software or something else?

We would prefer to use a different, campus-wide security system. We learned through the project that there are several Cornell Internet domains, including in addition to the main domain in Ithaca, domains for the medical school in New York City and the Cornell-in-Washington program. Adding these IP addresses to the server was awkward; a campus-wide security system would be preferable.

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of skill-sets they needed to effectively accomplish these. Estimate the amount of time each person spent on this phase, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

COMMENTS:

Log Files

Did you maintain log files of the use of MESL data?

Yes.

Did you perform regular analysis of those files? If so, describe.

No regular analysis was performed.

Did you produce regular reports at intervals throughout the project.

No.

Please identify contact person for information about log files. (title, email, telephone).

Noni Korf Vidal, Curator of Visual and Electronic Collections, nk11@cornell.edu, 607/255-3530.

Software

If you analyzed log files and/or created reports, what software did you use for these functions? Did you use off-the-shelf applications or build things from scratch yourself?

N/A

If you could start over today, would you use this software or something else?

N/A

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of skill-sets they needed to effectively accomplish these. Estimate the amount of time each person spent on this phase, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

Current MESL Support

This section is intended to provide insight from the MESL implementation side on requirements for support of MESL. Various user surveys are concurrently being conducted to give us the end-user perception and experience with MESL images.

Outreach

Describe any outreach, publicity or education activities you conducted to encourage the use of MESL at your institution

Outreach efforts followed two distinct paths. During the first year of the program, a general program intended to inform broadly the campus about MESL was implemented. The MESL Project Coordinator established the Cornell MESL Web site with general information about the MESL program, a FAQ about MESL at Cornell, links to the Getty MESL site, and contact information for further questions. The Project Manager distributed via e-mail to all deands, directors, and department heads (about 700 people) a description of the MESL project. Informational flyers were distributed to campus departments, and the MESL team sponsored a workshop on MESL at which the database was introduced to the campus community. MESL project staff described the project, and two faculty members spoke about how digital images could be used in teaching and their use of MESL. The workshop was held in a computer-training classroom with the intention that everyone would have a computer in front of them. There was an overflow crowd for the workshop, however, so some had to be content merely with the display. In addition to this campus-wide workshop, a demo of the MESL database was also conducted for the Art Department faculty at a faculty meeting.

The general outreach efforts identified a core group of faculty interested in using MESL in their classes. One-on-one involvement with these faculty represents the second MESL outreach path. The MESL project coordinator worked closely with the core faculty to develop subsets of data, customize the MESL interface for particular courses, produce course-specific pedagogical material, and in general make the MESL data useful to teaching.

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of events. Estimate the amount of time each person spent on this phase, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

Usage Instruction

What kind of assistance was needed for classroom setup and access, educating instructors on usage, helping students, etc. Please list and describe the kinds of activities.

Training needs changed during the course of the project. At the start of the project, faculty requested that the MESL Project Coordinator conduct training classes for the students. Because the delivery method of choice was the Web, with which most students were familiar, there was little need for this training, and it was dropped during the second year. There is no evidence that the lack of training had a negative impact on student use of MESL. Training was also offered to students in one other course, where MESL was used in conjunction with another, non-Web, Cornell database. The purpose of this training was to teach the students how to integrate their use of both databases.

Demonstrations of the MESL database were given to individual faculty. During the course of the project, faculty knowledge of and interest in digital collections and delivery mechanisms increased dramatically.

Classroom setup and access were the responsibility of facilities staff in the buildings where MESL was used for teaching. The MESL project did not require any different procedures or equipment.

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of events. Estimate the amount of time each person spent on this phase, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

Technical Development

Describe the kinds of technical development and support you offered to those using MESL content. Include activities related to the development of new MESL-related tools, incorporating changes into your MESL search or user interface functions, etc.

As noted above, one of the most important functions performed by staff was the customization of the MESL data for faculty. This could extend from merely preparing a subset of images for a browsing collection to building an online gallery of selected images.

People, Skills, Time, Cost

List the staff or positions involved in this stage of the project. Make a very short list of skill-sets they needed to effectively accomplish these. Estimate the amount of time each person spent on this phase, and include a "confidence level" to your estimate (Levels ranging from A for a wild guess to F for careful record-keeping).

Comments:

Support for Ongoing Use

This section is intended to provide speculative insight as to future needs in order to continue to support MESL use.

Activity Description

Describe the kind of ongoing outreach, usage instruction, and human technical support you need to offer for those making future usage of MESL content. Include activities and support related to instructional facilities, instructor and student assistance and training, and technical development such as changes into your MESL search or user interface functions, etc.

Some of the technical enhancements we would like to see in future MESL implementations are outlined in section 2.1.4.4 above. Improvement of the visual browsing capabilities of the web interface is of highest priority. Also very important is some standardization of the vocabulary used to index the images. A repeated criticism from users is that the different indexing standards in such a heterogeneous database makes effective searching almost impossible. The addition of full Boolean searching to the MESL search system is also a high priority. Finally, we would like to find a way to transfer the customization of the interface to the individual faculty members (or their graduate students), so that we no longer have to customize the MESL interface for them.

People, Skills, Time, Cost

List the staff or positions that would be needed for this. Make a very short list of skill-sets they needed to effectively accomplish these. Estimate the amount of time each person might spend.

Please comments and provide explanations and justifications for each position.

The Web programmer's primary responsibility would be to improve the interface between the database and the Web. The database, for example, is capable of conducting a Boolean search; the programmer would develop a query screen that could pass a Boolean search to the database. In addition, the programmer would develop the visual browsing tools required by the faculty.

The catalogers would address the issues of inconsistent data entry by re-cataloging the data according to common standards (most likely the terms found in AAT and ULAN), supplemented with our locally-developed thesauri.

The faculty members (or more likely their graduate assistants) would be trained in web interface design so that they could create subsets of the MESL data as needed for individual courses.

Experiences

This section is intended to give participants an opportunity to reflect on the MESL technical experience. What was easy? What was difficult? Specific problems encountered? Questions unanswered? We will probably want to follow-up with structured interviews for this more qualitative portion of the assessment.

Description of MESL development process

Describe how you responded as a project team to accomplishing the MESL technical tasks

When the MESL project started, there was general agreement that MESL needed to be on the Web. Unfortunately, the tools available for delivering image databases via the web and the in-house expertise for the completion of such a project was limited. The MESL technical team met initially, therefore, to identify possible delivery options. Once a likely delivery option was selected, a team member then explored further the implications of developing that delivery option. Several options were pursued that, in retrospect, were counterproductive. Some time was spent experimenting with a student-designed and built database, but because there was no on-going support for the database, it was not a viable option. Other time was spent with a commercial product that we naively hoped would develop an interface to the web; instead it went out of business. Other experimental and commercial databases were suggested but after researching their ability it was found that campus support was lacking. The result was that a number of dead-ends were fully explored before the implemented solution was developed.

While the process was frustrating in its slowness, the willingness of team members with widely-divergent backgrounds and skills to communicate with one another made the process tolerable.

Once the preferred delivery model was established, a two-person team took over the development and maintenance of the MESL implementation.

Design issues

Describe the design process you used in delivering MESL data to users. Did you redesign during the course of the project?

Given that there were no models to follow for the effective delivery of visual information, the MESL team elected to build an interface following a library catalog model. The first key component of the initial design was a search screen that allowed users to search for terms in specific fields. Query results could be displayed with first a brief, then a long, view.

There was little redesign of the initial screen layout during the course of the MESL project, though there were continued small improvements made to the interface. For example, the initial implementation did not have a generic keyword search. After brief experience with the data provided by the museums, it became apparent that providing only fielded searching would not be sufficient. The museums were not consistent in the terms used in the data, nor in the fields in which similar data was stored. A general keyword search that allowed the broad searching of all indexed fields was therefore added. Another refinement was the addition of links to the MESL field definitions, allowing users to explore more fully the nature of the MESL data and to improve their searches.

If time permitted, it would be desirable to conduct a formal evaluation of the MESL interface and see how, based on the experience of users, the interface might be redesigned. Unfortunately, most of the MESL project time was spent first in mounting the data, and later in promoting its use and customizing the interface for users.

Image Processing

What types of problems did you encounter while processing the images? (e.g., missing images, corrupt files, not having the right software to convert the format that you received)

As with the other MESL participants, we encountered almost every image processing problem imaginable. Some images were missing from distributions, and others were misnamed. One museum changed the filenames of the images in the distribution during the second year, forcing us to redo work from the previous year. Another institution removed images from the second distribution without telling us; the copies of these images distributed during the first year remained on the server until the anomaly was brought to our attention. A third institution submitted digital images that included the mounting material for the image. At thumbnail size, the image was lost in the mounting. We cropped those images to remove the mounting.

Fortunately, software was not a problem, though we did have frequent hardware crashes as our machines attempted batch processing of large groups of images.

How does the image conversion process from the first year compare with that of the second year?

The second year was much easier, thanks to the experience we had gained with image processing in the first year.

Can you make any observations, generalizations about image quality based on your MESL experience?

Our best indication is that image quality for the users was not the problem we thought it might be. There is no indication that the images we provided were of too low a quality to be of use, or that much larger or higher-quality images would be of greater utility.

Text Processing

What types of problems did you encounter while processing the structured text data? (e.g. character set problems, missing fields, delimiter problems, corrupt files)

The initial distribution of data was problematic. Much effort was spent in cleaning up this data, only to have a re-distribution of the data wipe out the effort expended on the first set. Still, the data did become "cleaner" over time, though foreign characters presented problems throughout the life of the project.

One problem that was never solved was the inconsistent use of the MESL fields among institutions. Two museums might record the same type of data in two different fields; others would record different types of data in the same field. Dates were a particular problem. One institution might record a BEGIN DATE and leave the END DATE blank; another might record a BEGIN DATE and put a zero in the END DATE; a third might record only the END DATE. The result was a database that, while information rich, was difficult to search effectively because of the inconsistent indexing.

The fine granularity of much of the data also made presented problems. In particular, it made it difficult for users to uncover broad categories of information. For example, users could not find all African objects because the cataloging received from the museums did not distinguish objects at this broad level, but instead used finer distinctions, i.e., Yoruba or Nigeria.

How does the text conversion process from the first year compare with that of the second year?

As noted above, both the cleaner data and the experience we had gathered during the first year made the second year much easier.

Security Issues

Were you aware of any security problems with MESL data during the course of the project?

No.

How did this happen? How did you respond?

N/A

Physical Infrastructure

Are your campus physical facilities and technological infrastructure robust enough to support teaching with MESL Images?

Cornell has one of the most robust infrastructures of any University in the country. Nevertheless, the MESL experiment uncovered a number of infrastructure problems, including:

Please indicate any non-obvious specific functionality that you think is desirable in student labs or classrooms trying to make use of MESL images.

None comes to mind.

Do your instructors need better access to appropriate workstations - offices, laboratories, classrooms? Please explain.

Poor equipment among faculty was a continuing problem. One faculty member at the start of the project had a grayscale (not color) monitor. Another only had an 8-bit color monitor, limiting screen display to 256 colors. While faculty machines are improving, they still lag behind students in their access to the images.

Do your students need better access to appropriate workstations -laboratories, libraries, classrooms? Please explain.

Student access to high-end machines was very good. They were more likely than faculty to use the high-end machines in the laboratories and libraries, and many had their own machines that permitted access from dormitories.

Does your university have a current upgrade plan or path for MESL capable workstations (both hardware and software, laboratory, libraries, and classrooms)?

Upgrading equipment is the responsibility of each individual department. There is no campus-wide plan designed to keep students and faculty current on technology, though there are continuing efforts to improve the equipment in the laboratories.

How Difficult Was This?

How did MESL rank (in terms of difficulty) among other information technology projects your university has undertaken?

MESL presented a different set of problems than our other information technology projects. Previous experiments with digital visual collections had run into administrative hurdles, with unclear channels of authority and communication. The problems we faced with MESL were overwhelmingly technical: how to clean up the data; how to manipulate and store image collections; how to make the data and images available via the web; and how to structure the data and images to be of greatest use to faculty and students.

Did this project nearly stretch you to your limits?

The project did push the MESL staff in technical directions that were new to them, but probably not to their limits. Of course, some features we would have liked to try implementing were avoided because of their high degree of technical difficulty.

Would MESL have been much easier to begin 2 years later? Knowing what you do now, would you do this project again?

There is no question that MESL would be easier to undertake now than it was two years ago. In the last two years the physical infrastructure of the university has improved, and faculty members have more and better computers. They have also become more knowledgeable about the potential use of digital images in the classroom (though some of this is due to the existence of the MESL project). Software, tools, and computer languages have evolved.

Observations on usage

Do you have any thoughts about MESL usage at your University. This is a chance to discuss the various usage pathways (access to MESL including outreach, workstation setups and locations, etc.) and obstacles to use encountered during your MESL implementation. Offer plausible explanations to patterns given your local implementation (even speculation as to why these patterns occurred - system not available, classrooms not configured, not easily accessible, lack of knowledge, lack of relevant images, etc.). Suggest possible remedies to these issues (if any)

While the previous discussion in this section has highlighted some of the technical and intellectual barriers to use, the biggest single impediment to increased use of the MESL images was the limited nature of the content available via the MESL project. Faculty who would have been delighted to use MESL images in the courses were frozen out because either the images they needed were not to be found in the MESL database, or they were present in such low numbers so as to make the inclusion of MESL images as part of their teaching not worth the effort.

The only solution to this problem is to expand greatly the number of images available via MESL, and to encourage the active involvement of faculty in their selection.

While the MESL pilot may not have had the images faculty members wanted, it was also difficult to tell them what MESL did encompass. There needed to be much more education about what the images were that were included in MESL, and why faculty might like to include them. There was no general description we could share with faculty and students on the nature of each of the collections provided by the museums, the particular strengths of each selection of images, and the reason for their selection. Such MESL "packaging" or public relation information could have been of great use in our outreach efforts.

Scheduling was problematic. The academic semester is an immutable pressure which was not accommodated by the MESL delivery of data and images. Cornell's Fall semester begins toward the end of August. In neither year were the materials ready by that date. It became increasingly difficult to encourage new faculty users when the end of the project was drawing near.

Other Observations on MESL

Once the technical hurdles were surmounted, communication problems surfaced. In particular, we would have liked to have seen more communication among museums and universities over content selection and image access issues. The communication problems may have arisen from the clash of two different "communication cultures." E-mail communication was not seamless; responses to queries were sometimes not timely or were non-existent.

Are there individuals on the campus who have had or are having an interesting or insightful experience(s) with using mesl?

Could you provide a short description of activities, projects, and experience - and any other insight. These experiences could range from active users, through frustrations and even complete refusals,

If possible, could you provide names, phone number, and email addresses for each individual so we can contact them.