NINCH guide home interview table of contents previous interview next interview
HATII interviewed Wendy Marcus Gogel, the Digital Library Project Liaison at Harvard University Library, and Stephen Chapman, the Preservation Librarian for Digital Initiatives, on January 26 2001. The Office for Information Systems at Harvard University, conceived of the Library Digital Initiative as a means of creating a comprehensive infrastructure to manage the digitization activities of the University. In the creation of a union catalogue, the Initiative seeks to address issues of acquisition, organization, delivery and archiving of digital deliverables within the library, with the resultant material being aimed principally at the University community.
Harvard University launched the Library Digital Initiative (LDI) in July 1998 as a five-year program to develop the University’s capacity to manage digital information by building infrastructure and expertise in this area. The three main areas of program activity are: Advisory Services, Technical Development, and the Grant Program. One part of the program is based on an internal bidding process, the Internal Challenge Grant Program, open to all parts of the University. This decentralized model for building content follows Harvard’s tradition on building library collections, keeping decisions about what to collect at a level close to researchers and students. The program is managed under the direction of a Steering committee of representatives from faculties and libraries. This is responsible for setting priorities for program activities, for monitoring progress and costs, and for allocating the incentive fund. The selection process for the Internal Challenge Grant Program is managed by the Grant Review Committee. Each project that has been awarded a grant has its own project manager who collaborates with the central Library staff. Because of the different projects that are part of the LDI, there is variety in the specific digitization approaches followed, depending on type of material, staff involved, purpose, and particular circumstances. Therefore, although they act as both developers and advisers at the LDI, they are not a centralized authority. They have learned what people want to control and what they would like to pass on centrally (e.g. digital preservation) and are focusing on these areas, operating on a consensus basis.
The individual projects that are bidding for an internal LDI grant are encouraged to carry out a collection survey. The way this is carried out and the level of detail recorded vary between projects. In practice many projects based their methodology on a very small sample of the material. As a result, some work plans were based on insufficient information and additional analysis and work was required.
The LDI Grant Review Committee is responsible for setting general priorities through the Grant Program, while the LDI Steering Committee may set strategic initiatives, such as the access projects specifically intended to provide enhanced access to collections by adding information to Harvard's various catalogs. The requirements explained in the Call for Proposals for the Internal Challenge Fund specify that successful proposals require institutional support, collaboration with LDI staff, library sponsorship for proposals submitted by non-library units in order to ensure access to information, and project management from the project unit. The selection criteria are based on the nature of the materials, the expected level of use, the level of sponsor commitment and contribution, the contribution to the LDI collection and infrastructure, the evidence of user input in project conception, the project difficulty and the likelihood of success, and the collaboration across collections (http://hul.harvard.edu/ldi/html/current_call.html).
The initial proposal of November 1997 for the creation of the LDI stressed the importance of content, stating that ‘digital library developments must be driven by collection content and by relevance to the University’s academic mission, not by technology’ (http://hul.harvard.edu/ldi/html/ldi_origins.html). In practice, however, content and technological challenges were equally important. Infrastructure development depended upon the selection of a range of formats that would pose different technical challenges. They used the projects to learn from them, letting each content contributor decide for themselves about the materials to be digitized. There was only a general guideline that the material had to be of value to the research community. This might be changing to some extent, however, as in the fourth round of applications the Steering Committee decided to specify areas of content that they would like to target.
A general observation about selection of material for digitization is that money plays a major role in selection. Grant applicants are designed to find out what is more likely to receive funding, in order to select projects and material with that in mind. In the first round of grants, there was a very open selection process. In the second round, applicants compared with what had already been done and adjusted their proposals accordingly. The condition of the material is not evaluated as part of LDI, since preservation is beyond its funding scope.
The Library would like to build a critical mass of digitized material and has followed a decentralized model about the specific decisions taken. Currently, the Library is focusing more on exploring the issues related to the acquisition of material ‘born digital’. Because of the sheer volume of material and the importance of recent ‘born digital’ material to research, this is a top priority. By comparison, although the digitization of historical collections is important, it is not a key priority, especially since the ‘analog’ infrastructure for locating the material already exists and has served students and researchers successfully up to now.
Their experience working with different departments of the Harvard community highlighted the importance of honest communication at a very early stage between project staff and the LDI advisors and technical staff. Good knowledge of the material is key. The role and amount of time required for project management was initially underestimated. For example, project management was initially carried out by curators who needed to find time for red-flagging and problem-solving along with other full-time duties.
Some of the obstacles to planning the development of digital deliverables are: technological change and uncertainty; lack of standards; reliance on proprietary solutions for some aspects, but also lack of commercial solutions for others; need for general education around new issues and broad consensus building.
Obstacles to building digital deliverables are: lack of experience in libraries in managing projects in general, and digital projects specifically; the decentralized nature of the institution; the time required to educate the community about new issues and technologies; the greater demands these processes place on the resources (staff and budgetary) of already busy departments; the difficulty of supporting technically decentralized activities (i.e. reliance on many networks, systems, and programs to accomplish tasks makes diagnosing problems difficult and time-consuming).
Another lesson learned is the difficulty of “painting the boat while you are sailing it”, that is, building the technology at the same time as trying to fill it.
The Grant Review Committee uses various criteria to evaluate grant proposals in the context of all of the submittals in a round. The strengths and weaknesses of each project are considered for each of the areas listed below (ranked as having approximately equal importance) with the exception of intellectual property rights, which must be cleared for the intended use of the material in the project as a prerequisite for funding.
Also (not in our questionnaire list):
These changed over time, with some criteria being added to the overall program with the addition of strategic areas identified by the Steering Committee. The Committee identified areas that were not already addressed by grant funding. For example in the Round Four Call for Proposals for the Internal Challenge Fund in August 2000 a list of Desiderata was added which mentioned that “LDI has not undertaken work in the following areas yet, and projects that address these challenges may be favored for selection:
Three other initiatives are not part of grant funding, but deal with:
These are all a fairly recent developments. Prioritization criteria did not change substantially over time.
Although the main activities related to the program were carried out in-house, there was wide-ranging co-operation with other organizations including archives, other libraries, museums, academic institutions, corporations, foundations and charities, and historical societies. These spread geographically across the whole range from local to international.
This experience showed that in any collaborative project, it is important to get things right: share standards, agree on best practice, have common guidelines, and colleagues open to communication. All partners need to be very clear when starting a collaborative project or collaborating to build development program.
The purpose of the Digitization Projects is to build digital collections of obvious and persistent utility to the Harvard community and, in some cases, the scholarly community at large. Digitization is not undertaken for preservation purposes.
The mission of the Library Digital Initiative (LDI), the digitization program, is largely influenced by the challenges to acquire, manage and deliver content that is ‘born digital’. It was set up in July 1998 as a five-year program to develop the University’s capacity to manage digital information by:
A statement of intent was produced with the Proposal for the Harvard University Library Digital Initiative, which was explicit about the rationale, scope, significance, primary audience, and long-term sustainability of the program (http://hul.harvard.edu/ldi/html/ldi_origins.html).
The program digitized a very wide range of source material with varied format and nature:
The materials selected for digitization were both the entire body of some collections, as well as representative samples, depending on the project. For some projects the digital deliverables were intended to be re-used. For example, the Harvard University Art Museums/Fine Art Library grant project was funded to deliver digital images through an online catalog, but archival versions produced for the project will be used for museum publications. The Baker Library grant project was funded to deliver digital images through an online catalog, but are also used for an online exhibit, as well as a poster and announcement for a library exhibit of the original source material. (description of both grant projects under Round One at: http://hul.harvard.edu/ldi/html/fundedprojects.html.)
For representing content they used GIF for b/w page images and JPEG for grayscale/color page images and pictorial materials. For describing content, they used MARC, EAD, and a localized implementation of the Visual Resources Association Core Categories for many images. The Making of America II DTD was used for brief descriptive metadata and labels association with page-turned objects. The Content Standard for Digital Geospatial Metadata from the Federal Geographical Data Committee and the Draft of ISO19115 are being used to develop a system for finding and using GIS datasets. They use XML for representing structure and the Making of America II DTD for representing the structure of page-turned objects. For sound files, the AIFF format includes structural metadata in the form of channel definitions and time marks.
Because of the decentralized organization of the Library, there are several different systems behind the scenes. There is a need to build a system that works. This is the incentive for individual contributors to make compromises.
For controlling data values they use AAT, Library of Congress subject headings, name authority files, and MeSH, Library of Congress. They use standard vocabularies whenever possible, but they cannot dictate that various projects use the same vocabularies, hence the use of AAT, MeSH, LCSH, etc., and of LC Name Authority File, Union List of Artists Names, and others.
They looked at existing guidelines for digitizing particular document and object types, but did not adopt any (apart from the Ameritech/LC and the MESL ones for projects they participated in). For image and text digitization, rather than prescribe specifications for classes of materials, they developed procedures that create project-specific specifications. The goal is to produce digital reproductions that meet but do not exceed near- and long-term functional requirements (http://hul.harvard.edu/ldi/html/reformatting.html).
In the case of audio digitization, their choice of archival formats relates in part to the systems and expertise resident in Harvard’s Eda Kuhn Music Library Audio Preservation Studio.
In the case of biomedical images, their choice of archival formats (TIFF, grayscale, palette color, RGB, deposited with or without additional Look Up Tables) was influenced by existing practices among scientists and researchers routinely creating digital objects in their labs. The desire is to accommodate practice rather than mandate changes.
They realized that one specification could not work for all users. The technical specifications vary according to source matter.
The primary intended audience for the digital deliverables was described as follows:
“The purpose of the Digitization Projects is to build digital collections of obvious and persistent utility to the Harvard community and, in some cases, the scholarly community at large. … Sponsoring departments for individual projects will be expected to … commit to on-going support to ensure that content made available under the program will continue to be available to the Harvard community …” (http://hul.harvard.edu/ldi/html/current_call.html).
More specifically the targeted audience is: four-year college, graduate school, lifelong, distance, and computer-mediated learning, and archive users.
The vast majority of LDI-related resources are accessible to all, so apart from the targeted audience, anyone who desires to use Harvard’s public catalogs and the linked objects that have not been restricted to the Harvard community, can do so.
They tried to acknowledge the needs of those with disabilities. They took ADA into account in user interface design, and followed locally developed guidelines for compliance with these regulations.
Licensed electronic resources are used a great deal. The new digital content created as part of the LDI Internal Challenge Grant Program has not reached the critical mass that is needed to evaluate its use.
In terms of evaluation of program requirements, generally the Harvard LDI system includes consensus of the community (through working groups, focus groups, open meetings) and oversight by committees.
Each project has its own project manager. Wendy Marcus Gogel is the project liaison for the central Library for the LDI projects. Project managers are required to provide quarterly updates. The LDI Steering Committee has the program oversight. Other Committees have the oversight for individual systems and/or projects. Project management procedures are project specific and vary considerably. LDI staff have project management expertise and offer advice to individual projects. Templates and guidelines are also available on the web. They always ask in the applications for scheduling, workflow analysis and user needs analysis. Workflow piloting is often carried out, too, mainly on a small scale. They might also request specification of training needs, technical feasibility, workflow analysis, workflow piloting, technology forecasting (whether the project will require new systems or not), and reformatting specifications. Generally staff do not have time for extensive planning. They often want to jump in and do it, but the importance of thorough planning cannot be overemphasized.
Digitization was both carried out in-house (vast majority), purchasing the equipment, as well as being outsourced (35mm microfilm digitization). Preference for in-house digitization was due to the main emphasis of the program on developing local infrastructure and skills. Other important motivating factors were the opportunities for bundled services – materials preparation, digitization, metadata creation (structural and administrative), and loading of data to Harvard’s Digital Repository Services. When selecting the outsourcing option, they warn that ‘significant time must also be budgeted for other activities: materials preparation and transfer, digital image processing and quality control, structural and administrative metadata creation, file management, and storage’ (Chapman & Comstock 2000).
|
Technologies Used for Image Digitization |
|
Flatbed scanners |
|
Film scanners |
|
Digital cameras |
|
High-end professional cameras |
Their experience showed that operator skill is the most important factor in the use of digitization equipment. They established a set of guidelines for document scanning and digital photography. Gray scales, color charts, MTF slanted edge targets, and the Macbeth ColorChecker were used as benchmarks. Some of these were used to generate the ICC profiles. For archival copies the tone and color reproduction was done to the target, not the source file. Detailed descriptions of the Digital Imaging Production Services at the Harvard College Library, specifying the preparation of the facility, staffing, equipment, workflows and specifications are included in the article by Chapman & Comstock (2000).
DIG photographer Stephanie Mitchell at the digital camera workstation. An array of the technical targets they used to characterize our systems are evident in the foreground (from Chapman & Comstock 2000).
Apart from the people who work on the various projects, the number of staff who work on the LDI program are as follows:
|
Type of Staff |
Number |
% of time on the project |
|
Manger |
1 |
100 |
|
Metadata specialist |
1 |
100 |
|
Technical development staff |
6 |
100 |
|
Projects liaison |
1 |
100 |
|
Project staff throughout the University (cataloging, technical processing, programming) |
8 |
100 |
|
Reformatting advisor |
1 |
100 |
|
Digital acquisitions coordinator |
1.5 |
100 |
The background and profile of most people on the central LDI team is technical. The remainder have a combination of library background and ICT. Staff were re-deployed from other areas. Although advice was available in-house about technical aspects of digitization, they used external consultants for space planning, studio lighting, and training in the use of specialized benchmarks. An area where training needs have been identified for the main LDI team is project management, where they will organize training in the future. Other areas include application of technical standards, preparation and handling of materials for digitization, post digitization processes (especially color management), and metadata creation. Training was received by the program director, library/cataloging staff, and equipment operators. This was organized in a variety of ways: in-house, using project staff, the Library’s own consultants, and external consultants, by attending external courses, with independent study, and by learning on the job. The training did meet the needs of the organization. In-house training is provided to use systems developed by the Harvard University Library. On the job learning and training from external consultants is solicited to solve specific problems/challenges – such as metadata extraction from Kodak Photo-CD images – that arise in specific projects.
Clearance of intellectual property rights is a pre-requisite for funding the projects. Without that, no project is approved. Users can download a variety of digital deliverables. No electronic management systems are used to control copying.
There is a conservation procedure for original materials. The investigations carried out into the condition of the original materials prior to digitization vary according to project: condition assessment, surface cleaning, re-attachment of photographs to album pages. No conservation activities are undertaken as part of the LDI program, since this is outside its scope. Human and environmental risks were identified during the preparation or digitization process. Human risks include materials packaging and transport, and handling during the process of digitization. Environmental risks include temperature and humidity, dust, lighting, and security. The risk was assessed by the Library Conservation Team which offers advice, training, and treatment. The material does not go by default to the Conservation Team prior to digitization. Risk assessment is also carried out with external consultants as part of space planning in the development of digitization studios. Site visits by the LDI Reformatting Adviser are carried out in consultation with conservators. They use a calibrated light meter to measure lighting intensity and UV. More information is available at the DigiNews article (‘Digital Imaging Production Services at the Harvard College Library’ RLG DigiNews 4(6), December 15, 2000). Steps taken to minimize the risk include the development of policies for materials transport, security and handling, and upgrade of equipment (e.g. they ordered UV filters for digital camera lighting in one studio). They also use custom-built cradles for photograph albums; they digitize all flat pictorial items under glass at the digital camera in the HCL (Harvard College Library) Digital Imaging Group. The materials were usually prepared by curators and sometimes by preservation staff. Curatorial or preservation staff do not normally prepare/monitor materials during digitization, but photographers/scanning technicians receive materials handling training from conservators in the Weissman Preservation Center and HCL Conservation Services. There are no plans to restrict access to originals once the material has been digitized.
The catalogs used at Harvard before digitization were HOLLIS, VIA and OASIS. The information used from these systems in the digitization process was the unique ID for each item. Access to all the relevant materials (e.g. cataloging information) for the digital deliverables varied according to project. Full cataloging is not required prior to digitization. Unique materials, such as manuscripts and photographic prints, were not altered from the original form for the digitization process. Some items received minor conservation treatment (surface cleaning, reattachment of photographic prints to album pages). Others were rehoused in mylar sleeves with attached labels to ensure that associations between digital files (filenames) and unique IDs for the source items were always correct. Non-unique materials are sometimes photocopied or disbound prior to scanning to facilitate autofeed production of 1-bit page images. No material was rejected before digital imaging. They digitized both originals, as well as using reproductions and intermediaries. For images they used photocopies, 35 mm slides, 4 x 5 or 8x10 transparencies, and microfilm. The form of intermediaries used for audio included 33 rpm discs, audio cassettes, DAT tapes.
The originals are generally cataloged. Often the digital materials are not cataloged separately, but are used to represent the original (e.g. for museum objects) or are described as formats, versions, or surrogates of the original (some visual images, page-turned objects). Some of the materials are not cataloged per se, but are described as the foundation of the image. In other cases, such as biological type specimens, the original specimen is the cataloged object, and digital images are treated as representations of it. The metadata standards used to catalog originals depend on the kind of material. AACR2/MARC cataloging, archival finding aids, VRA Core records, and purely local schemes are all used in various circumstances. They frequently add a holding for the digital object to a MARC record for the original. However, in some cases it is not necessary to catalog the digital objects as such. While they do generate and maintain technical and administrative metadata about the digital objects, there may be nothing they want to tell the user about the digital object itself other than that it exists as a representation of the original, cataloged object. In these cases there is simply a link from the description of the original to the digital representation of it. OLIVIA uses controlled vocabulary in several fields. The curators decide what thesaurus to use to populate the controlled vocabulary. The metadata records information about the original object, the digital object, the digitization process, technical details, and administrative information. The metadata records were created by a range of staff, including digitizers, archivists/information professionals, or digitizers who are also archivists/information professionals. Descriptive metadata records for digital deliverables are rarely created as independent entities. To the extent that the digital version is described, that description is available in the same context as the description of the original, whether an AACR/MARC record, an archival finding aid, a VRA-like image record, etc. These different metadata types are served by separate catalogs, but the distinctions between catalogs are not based on the difference between digital and analog deliverables. The catalog for the digital deliverables is available on the internet.
The format chosen for retroconverted text-based digital deliverables was XML markup. They used the Making of America II DTD to encode OCR-generated text that points to corresponding page images (stored as .TIF, delivered as .GIF). The documents did not contain non-Latin characters. The OCR software used was Prime Recognition (system in place at University of Michigan Digital Library Production Services). The level of accuracy achieved was 96% search accuracy (as opposed to character accuracy, which was not counted). They looked at bit-maps, typed in key terms, and checked the system for accuracy. Prior to the OCR process they used TMS Sequioa’s ScanFix software to optimize 1-bit page images. The aim of using OCR was automatic indexing. They did not employ any keying in.
The advice they would give to other organizations about to start using OCR would be:
They appreciated the University of Michigan DLPS’s service to join words separated by line-break hyphens.
The file formats used for images were GIF and JPEG for delivering (and PDF for page images), and TIFF, Photo-CD and RAW for capturing and preserving. The resolution generally ranges from 300-600dpi for capturing and preserving and 800 pixels (dpi varies) for delivering images. They used 1 to 48 bit-depth for capturing and preserving, and 4 to 24 for delivering. Group 4 and ImagePac PCD compression was used at capturing and delivering stage while JPEG compression, GIF/LZW was used for delivering. The aim of the compression was to reduce cost (by reducing file size, since storage in Harvard’s Digital Repository is charged annually by GB), to improve access, to enhance usability, and to decrease storage requirements. Users of the Digital Repository Services are informed about their options (e.g. save money by using compression, but less safe for preservation) and are left to make their own choice. The original scans are retained in uncompressed form, except for 1-bit scanning and Photo-CD scanning projects). Some post-scanning processing is carried out on the images. (For details see Chapman & Comstock 2000.) The average file sizes for grayscale and color images vary from 4-18MB at capturing and preserving stage, and 200 KB at delivery (Museum objects may be scanned at up to 150MB). They use specialized targets and software to measure performance of digital cameras and scanners. These same targets are used to generate ICC (International Color Consortium) profiles for color management.
Their experience showed that collaboration is necessary to arrive at commonly used ways of defining and measuring quality. They do not recommend prescribing digitization methods and technical specifications per class of original material (e.g. color photographic prints).
The documentation of the methodology and procedures they used to digitize sound is currently in process. The file formats used are AIFF for capturing and preserving and RealAudio for Delivering. They also use Sonic AIFF for capturing. The sampling rate used is 88.2 kHz for capturing and preserving and a variety depending on bandwidth negotiation for delivery. The bit-rate is 24 for capturing and preserving and 16 for delivering. The file sizes created were approximately 30 MB per minute at capturing stage and 1MB/min at delivery.
The program does not digitize moving images.
For the material deposited in the Digital Repository (about 90%) they use both automated and manual procedures checking for the quality of data and metadata. They used a variety of procedures for ensuring the quality of the digital deliverables: spot checks (human); check on a stratified random sample; percentage checks on carriers (e.g. corrupted disks, etc.)(automatic); total check (automatic); and check using MD5 checksums. For metadata recording of the LDI projects it works well when the QA is carried out by the project developers themselves, since they know their material best and can follow the LDI guidelines and keep communicating with the LDI staff. Procedures vary from project to project. For example, a cataloger may have format-specific knowledge, such as how to catalog images, but lack specialized domain knowledge such as Chinese culture or botany . The metadata records and linked images are reviewed by a domain specialist both for metadata content and for accuracy of image links. For some systems, new vocabulary terms are vetted by database maintenance specialists.
Quality control procedures may change the sequence of processing, generate systems requirements for particular projects (e.g. periodic lists of new terms, draft records plus thumbnails), or introduce new levels of review.
Harvard provides two levels of access to digital resources: some are open to the public, while others are restricted to the Harvard community.
An important development for the LDI is the Name Resolution Service (NRS), which provides a robust means of identifying and finding digital materials. This is a comprehensive service for creating, maintaining, and resolving names, which are persistent, location-independent identifiers for network-accessible resources. Name resolution is the process of mapping from a given abstract name to an URL that represents a particular instantiation of the named resource. The NRS is comprised of two separate sets of services and servers: (1) an Oracle-based administrative service that manages the metadata necessary to perform naming services; and (2) an HTTP-based resolution service that performs naming resolution (http://hul.harvard.edu/ldi/html/storage_access.html). More information about the infrastructure of the LDI is available on Dale Flecker’s article at D-Lib (Flecker 2000).
The methods used for informing potential users about the digital deliverables created are project specific with shared responsibility between LDI and the department/unit that is undertaking the project. It may include any of the following methods: announcement on website, press release, articles in print media, print and broadcast media coverage, announcements at conferences and meetings, announcement at electronic listservs, conventional mail shots, or registering with a web search engine. The LDI website includes for example, a page with information about news coverage of the LDI and pointers to the relevant articles http://hul.harvard.edu/ldi/html/ldi_news.html and a page announcing new projects funded http://hul.harvard.edu/ldi/html/funded_projects.html.
LDI can contribute in this area with its vision of being able to “expose” metadata to search engines at large from all types of Harvard sites to assist data harvesting in “deep” web space. They are currently working on the technical development for this. The idea is that users would be able to go to one of the Union catalogs, do a search, and have dynamic websites generated.
No front-end evaluation has been carried out. Formative evaluation was carried out for some projects using paper and online questionnaires, email, and focus group discussions. Some project proposals based their goals on the results of evaluation. For the LDI program as a whole, LDI is committed to a third year evaluation that is part of the original proposal. For individual grant projects, evaluation is required after project completion.
The full five-year program is funded at $12 million from internal Harvard University funds ($7million is for staff, hardware, software, space, overheads and $5 million is the incentive fund). They received everything they asked for.
All elements of the project will need to be updated over time. New materials and metadata will need to be digitized, metadata will need changing, the user interface will need to be changed, and it will be necessary to change file formats. They have a preservation strategy, outlined in the Guide to Preservation Services, that will be made public around March/April. This is the same for all types of material (text, images, audio); for example, file formats will need to be monitored for all media types, storage media and conditions for all media types, and metadata updated for all three categories. The strategy is based on migration of data. Quality control procedures in the life-cycle management are based on monitoring and migration. Longer-term sustainability is not dependent on self-generating funds, but on operational ones.
As general comments, they mentioned that communication is essential, due to the collaborative nature of the program. Planning and documentation are also very important. Their general advice is: ‘once there is agreement, document it’. What Harvard can contribute to the Good Practice Guide (which resembles the research undertaken by the HATII team for the Good Practice Guide), is the recommendation to talk to different players and record the lessons learnt. The effort invested in this shows that mediation and facilitation are crucial and support decision-making. Recipes of how to do things can only take you so far. This highlights the important distinction between getting things done versus getting things right.
Some interesting issues and points arise from this case. One is the decentralized model for content development that Harvard has adopted and how this is adapted in the digital environment: they give a large degree of freedom and flexibility in selection of material to a wide variety of contributors, ensuring rich content. They centralize those areas that nobody else wants to control, such as digital preservation and technical development. They place great emphasis on good infrastructure and architecture of the system. Their experience with different projects highlighted several common pitfalls, such as planning digitization projects based on incomplete inventories of collection, or the general reluctance of people to plan – an area they think is crucial. It is also worth noting that despite their efforts in conversion projects, the Library as a whole is giving a primary role to ‘born digital’ material. Finally, it is interesting to note that most of the projects that are part of the LDI program are add-on projects for the contributors, and this has subsequent project management and resource implications.
Chapman, Stephen & Comstock, William, 2000. ‘Digital Imaging Production Services at the Harvard College Library’ RLG DigiNews 4(6), December 15, 2000.
Flecker, Dale. 2000. ‘Harvard’s Library Digital Initiative: Building the First Generation Digital Library Infrastructure’ D-Lib Magazine 6 (11) November 2000.