While this was still early in the era of the Internet for everyone, it is helpful to consider why museums, along with archives, were still relative cultural institution latecomers to the online world. Libraries had had telnet and gopher access to OPACs for years by this point. Several explanations come to mind for why museums came on board only with the advent of the Web. For instance, previous Internet formats were limited to text; suitable to much library material, but inadequate for describing or effectively using museum collections comprised of non-textual artifacts. The Web changed that by allowing text and images to be delivered side-by-side. Perhaps more important, however, were the functional and cultural differences between museums and libraries. Museums still are competitive with each other within a given region and topic; two modern art museums within one city are in fact competing for that city populace's leisure time and dollars in the very direct sense that the dollars come in through door-ticket sales. Museums are often considered leisure time or tourist activities, meaning they are optional, and not a place one goes because one has to for research or homework as is the case with many libraries (especially university and research libraries). These direct forms of competition over funders (visitors) who are not compelled to spread the wealth equally, combined with the lack of clear opportunities for copy-cataloging, creates institutions in which collaboration and information sharing is not absent, but is less a priority than in the library community.

Perhaps one should ask why museums came on board at all, as they have recently with so many projects to provide broad, online access to collections. If you compare the traditional use of museum collections and library collections, you would find that libraries have collections which are, for the most part, directly and physically accessible to their users and in far greater volume than library staff could accommodate through direct mediation. So, it is natural that libraries developed access systems that visitors could use directly (card catalog, OPAC) and which in turn availed visitors direct access to the collections themselves. Museums, on the other hand, still operate under a mode of indirect, mediated access. Visitors are not encouraged or even allowed direct access to the collections, so no self-directed access system was needed for visitors. Rather, museum staff pre-select and present, usually with prodigious contextualization, a portion of the museum collections in the form of exhibitions. The exhibition is a melding of physical and intellectual access to collections. Given this traditional (and still dominant) mode of access for museums, along with the factors mentioned above, it is somewhat a mystery why museums on the whole have devoted so much energy in recent years to individual institutional websites, and to large-scale consortium projects which are mainly collections access oriented (not nearly as many regional or national collaborative projects exist for the obvious museum functions of exhibition and education). It should be noted, however, that museums always have, and continue to provide direct collections access for research purposes, but it has remained, until now, a somewhat under-promoted and secretive mode of access in comparison to exhibitions.

Collections data has always been the most rigorously structured and tracked information in museums, and this perhaps lends itself to computerized access systems and gives museums a natural starting point which, when combined with the Web's ability to include images of objects and reach more of those sought-after visitors, provides a basic explanation about the timing of museums' entrance to the world of online access. Despite this quick and admittedly speculative consideration, these historical and museological topics have not been rigorously researched enough, and are sure to provide insights into the future of cross-community online projects. Perhaps these consideration will provide this report a starting point by introducing some of the institutional motivations related to online access that inform a cross-community project like MOAC. In turn, the history and development of MOAC may help illuminate some of these issues.

Why MOAC? Rationale and Reason for Being

It has become increasingly clear that researchers, educators, students, and even the general public are expecting as a matter of course that society's public institutions should be informationally open and accessible. It is unclear whether supply or demand has driven this expectation more, but both have combined to create a culture of opportunity and heightened expectations.

On the demand side, this social shift surely is in part motivated by the thrust to hold public institutions, especially government, more accountable as evidenced by the Freedom of Information Act in the U.S. It might be also be driven by the development of media and technology in general. For instance, if shopping and interactions with commercial institutions become quicker and more convenient because of the Internet, then why should the public accept much more arcane and difficult modes of access to publicly held information? The public won't accept this, and subsequently updated the Freedom of Information Act to include a new Internet component.

On the supply side, public, governmental, cultural and research institutions such as museums, archives, historical societies, universities and libraries have seen and pursued opportunities to significantly expand and improve some of their core functions such as facilitating research or furthering public education.

As exciting as these opportunities were for some cultural institutions, it was perhaps equally surprising for them to find themselves on a newly leveled playing field. Since so many of these institutions were now within easy reach of the public, the public need not care so much about where an individual artifact or book was held as long as they had some form of intellectual access to it. The researcher never did care as much as the cultural heritage professional what accidents of history gave rise to any one institution's collection; the researcher just needed to find what was relevant to their work, no matter whether it was a painting held by an archive or a manuscript held by a museum, and no matter where it was located. Loyalty to the museum or archive which previously had an exclusive relationship with a user-community based on access of proximity has waned. Museums, libraries, and archives, in this very particular sense, are becoming more important as a community than as individual institutions. Not to overstate the situation, regionally based museums and archives with primary physical collections hardly have reason to disappear, but for some informational, instructional, and research purposes surrogate representations of cultural collections such as found in textbooks, slides and digital images are entirely sufficient and are not regionally bound.

One key lesson for cultural institutions is that they should not expect their visitors and researchers to behave in a networked environment in the same way they did before networked access or that they do during physical visits. There is no reason now why a researcher should be satisfied at having to approach at the level of the institution: to divine the URL and contents of each individual institution website separately, learning new interfaces, new search vocabularies, and then collating the disparate information together. For many purposes, from research to instruction, visitors want to approach at the level of content, finding closely connected, if not federated, access to similar content from many institutions easily. This is true enough now of researchers seeking collections information, and will probably become more true of even less research-equipped audiences such as teachers and the general public seeking other types of content.

While such observations might seem obvious today; they were not so clear in early 1995 when most museums had no website and when 10 museums in California decided to leap over the institutional gap and work with a couple of dozen archives on a project we named MOAC (Museums and the Online Archive of California) to explore whether and how museums and archives might provide standardized integrated online access to collections.

"...standardized integrated online access to collections", "...more important as a community than as individual institutions". MOAC was on the right track, and had a lot to learn.

Who is MOAC? Organization and Business Model

1995 was also the pivotal year of the Berkeley FindAid Project that gave rise to an emerging SGML standard for describing archival collections - the EAD (Encoded Archival Description). This author attended an EAD symposium held by the UC Berkeley Library in March 1995 and began testing application of the EAD to Berkeley Art Museum collections such as conceptual art collections, which are combinations of artifacts and primary documents. This application of the EAD to primary museum collections would bring into sharp focus discussions about the role of standards in relation to community descriptive practice as detailed later in this report. Shortly thereafter, the University of California initiated the UC-EAD project, a prototype project to test the application and delivery of EAD encoded finding aids from archives, library special collections, and historical societies. In 1997, this author approached the organizers of UC-EAD and proposed that the project then included almost every type of cultural collecting institution with the exception of museums and that on opportunity existed to build a comprehensive cross-community resource. After hearing about early experiments using the EAD in museum settings, Daniel Pitti and other project coordinators agreed to a related prototype project which would explore if and how to bring museums into the UC-EAD project. In 1997-98, the UC-EAD project took 3 significant steps toward expansion. With funding from the California State Library, UC-EAD changed its name to the more inclusive OAC (Online Archive of California) and engaged many non-university partners. Management of the OAC was moved from the UC Berkeley campus to the more neutral and central location of the CDL (California Digital Library) in the University of California, Office of the President in Oakland, where Robin Chandler remains the manager of the OAC. The related museum project was named MOAC (Museums and the Online Archive of California) and 10 initial partners were selected to start working in partnership with the OAC in earnest.

The selection of the initial MOAC partners was carefully considered. The initial group was kept small enough to enable the project to move and scale quickly enough to achieve results within a couple of years. The idea was in part to develop a content testbed, but mainly as a means to an end. The goal of this group was to test the theoretical and practical conditions of museum participation in the OAC. If potential problems could be identified and solved, this project would scale up to more museum partners. In order to conduct a realistic test; we chose a representative selection of museums which included university and non-university museums, large and small museums, and museums covering a range of disciplines from art to anthropology to photography to ethnic history and beyond. Also included in the project was of course the OAC as a primary partner and one archive, the Bancroft Library, in part because the Bancroft was documenting a collection of paintings, but more importantly that the set of museums could work with the OAC and Bancroft to figure out how to bridge the institutional gap between museums and archives.

In 1998, Richard Rinehart and Tim Hoyer wrote a successful grant application, which was submitted by the California Digital Library to the Institute of Museum and Library Services, National Leadership Program, to fund the project from 1999-2001. The project was later awarded a one-year, no-cost extension, taking it through December 2002. The organizational model was that of a basic non-profit consortium. Richard Rinehart became project manager, and each participating institution named one or more working liaisons to the project Steering Committee. The list of institutional partners and Steering Committee members included:

Berkeley Art Museum/Pacific Film Archive - Richard Rinehart, Guenter Waibel

Japanese American National Museum - Cameron Trowbridge, Snowden Becker

Bancroft Library - Mary Elings, Eva Garcelon, Tim Hoyer, James Eason

Online Archive of California - Robin Chandler

Cantor Art Center - Leslie Johnston

California Museum of Photography - Steve Thomas, Jennifer Frias

Oakland Museum of CA - Dana Neitzel, Lori Lindberg

Grunwald Center - Layna White

Hearst Museum of Anthropology - Josh Meehan

Fowler Museum - Sarah Kennington

[The Museum of Paleontology and San Francisco Museum of Modern Art would join later.]

The project established an internal email list for ongoing work and communication, as well as a website which both documented and disseminated the project development, as well as providing access to the growing content resource. In the latter sense, the MOAC website is a 'content portal' into the larger resource of the OAC; so searching the MOAC website will reveal only MOAC resources from within the OAC, but searching the main OAC homepage will reveal MOAC content alongside all other OAC content. The content portal was devised after it became clear that the MOAC content had more item-level metadata than the average OAC finding aid. Since this metadata did not exist in most OAC finding aids, it was not searchable via the main OAC interface. This metadata (artist, item title, etc) was considered important in discovery and presentation for MOAC content; and the MOAC portal provided a way to achieve enhanced access to specific content within the OAC, while retaining the benefit of having the content integrated with the rest of the OAC. Afterward, other OAC related projects, such as the Japanese American Relocation Digital Archives, created similar portals or customized views into specific subsets of content within the larger OAC.

The business model for MOAC was set up as, in short, a public works project, with all the unapologetic idealism that implies. This is not to say that the partners in MOAC have been fiscally na•ve or careless about practical issues; in fact they have been quite the opposite, and have considered at every turn the very practical obstacles that would face small museums hoping to participate in MOAC, including economic, infrastructural and expertise limitations. MOAC partners also did not reject outright the important experiments that were starting at the same time that involved industry partnerships and semi-commercial business models. Rather the members of MOAC decided that there was room in world of digital library projects for testing a model in which the content and direction came directly from a broad base of public institutions (museums, universities, archives) with a strong mandate for public service, and would endeavor to serve directly the greatest possible public good with free and broad access. Commercial incentives and partnerships could arise from secondary activities (such as licensing online images for re-use in commercial publications), but that such activities should not interfere with the primary goals of broad participation, broad dissemination, and public service. This project would explore and affirm the model of the public sector as important to education, the emerging Internet, and society in general.

What MOAC Set Out to Learn

MOAC established a focused initial research agenda of practical and theoretical issues that has expanded over the last 3 years as a result of unforeseen developments in technology and collaborative opportunity. MOAC would measure success on the ability to answer these questions that arose from the research agenda, and not necessarily on the ability to prove previous assumptions. That agenda included the following areas of investigation.

MOAC naturally investigated how best to integrate collections information from both museums and archives, as well as library special collections, historical societies, and other organizations participating in the OAC. The goal was to create a comprehensive cultural heritage resource that respected institutional differences, but served the end-user who wanted to access content that crossed institutional and even community boundaries. If the Hans Hofmann painting collection was at the Berkeley Art Museum, and the Hofmann Papers at the Bancroft Library, researchers might be well served with the ability to discover both of those with one search. Beyond specific content queries, much research and education is serendipitous, and presenting so many types of collections in new combinations might also spur new scholarship.

MOAC set out to learn whether the emerging EAD standard could be appropriately and effectively applied to museum collections (including not just document or 'archival' collections in museums, but also to primary artifact collections), and in various real-world museum environments. MOAC chose to call their EAD encoded documents 'collection guides' rather than 'finding aids' because the term was clearer to museums and the EAD accommodated both types.

Since museums typically have collections data at the item level already, usually contained in a museum collection management system, and since images are so important to discovery and research of non-textual collections, it made sense that MOAC should also research how to implement the EAD at the item level, and how to link EAD item-level description to images and multimedia - both relatively new questions with regard to the then emerging EAD standard.

MOAC set out to investigate the implications of decentralized content creation and storage, enabling each partner to convert their own data into standardized formats, and to contribute some content, such as EAD collection guides, to a central server for cross-searching purposes, while storing other content, such as images, on local servers where they were linked from within the collection guides. Part of the reason for this investigation was that partial decentralization might allow for quicker and larger scalability of the project by distributing the resource load. The other intention was to use the opportunity of the MOAC project to 'bootstrap' and help enable museums to build their capabilities for the long term, instead of concentrating exclusively on building the content portal.

MOAC was a testbed created, in part, to develop adaptable methods of content creation and delivery that could be replicated by any museum in California. So, MOAC also set out to research and recommend cost-effective practices which dovetailed with existing typical museum workflows, that took into account typical in-house museum expertise with standards and new tools, and that relied on data which was extant in a typical museum's collection management system, requiring no new cataloging.

What MOAC has learned

The basic definition of a collection as defined by the descriptive practices of the museum and archival communities informed both questions about the appropriateness of the EAD for museums and the ability to integrate museum and archival data. Archival practice defines a collection as the unself-conscious by-product of the activities of a person or organization. That collection comes to the archive where it is kept in the original order of the previous collector and described in terms of that provenance and its contents. A certain amount of objectivity is desired in the description, which is limited to the scope and history of the collection. Archival collections are maintained with the integrity of the whole, also defined by provenance, and are described from the top collection level, and only sometimes at further levels of detail. Archival collections are typically paper-based records and documents. Museum collections are often acquired at, and oriented toward, the item level where most description tends to focus less on the archive-like history of the object, or the library-like subject of the object, and more on the 'thingness of the thing' including the physical properties such as material, dimension, and object or genre classification. If expanded cataloging can be afforded, it often includes details about the creator (not necessarily collector) of the artifact, as well as interpretive text. Those are the respective descriptive practices in theory anyway. Practices that at first glance seem almost irreconcilable in creating an integrated access system.

In real world archival and museum environments, descriptive practice is more muddied, and while that may cause intra-community problems, it may also create inter-community opportunities. Practical considerations often require archives to store collections not in original order, and intellectual access considerations often require description that does not strictly follow original order. There is a professional archival term, 'artificial collection', for the practice of organizing a collection, either physically or intellectually, in ways that are not based on the integrity of one collection from one collector kept in original order. Archives are often quite adept at interpreting their collections, and with the recent desire to digitally image collection items, archives are increasingly creating more item level information.

Museum practice is even more flexible, combining different descriptive techniques in one institution and even in one collection. For instance, some collections come to the museum, and are maintained, as a whole much like archival collections. These might also be described by their provenance. However, since museum collections are typically artifacts with heterogeneous storage demands, they are rarely maintained in the original order of the previous collector. Moreover, museum description of that collection may combine archival-like provenancial information with more interpretive text in which the museum creates new intellectual orderings, such as breaking the collection into media type or themes that were not created by the previous collector. Museums regularly group items that were acquired individually into what archives might call 'artificial collections' created by the museum, and ordered according to thematic or media type. However, while museum collection management systems typically contain only item level records, it would be a mistake to conclude that museums think of their collections at only the item level. Rather, most museums also create aggregate collections though this collection-level data is not stored in the collections system, but rather in more unstructured formats like gallery brochures. A quick visit to the websites of any mid-sized to large museum will reveal descriptions of specific sub-collections including provenance-based descriptions such as the Hofmann collection at the Berkeley Art Museum or the Panza collection at the Guggenheim Museum, and 'artificial' collections such as American Decorative Arts at the Met Museum. So, museums and archives engage in similar descriptive practices by creating both provenance and artificial collections; but in inverse proportions.

This confluence creates a friendlier environment for integrating museum and archival description, but it should be admitted that completely seamless integration might not be realistic. Rather each new project should weigh the benefits of integrating access with the potential for confusing end-users with different types of description in one system.

The implications of this difference in descriptive practice should not be taken lightly or dismissed out of hand. These issues deserve further attention in end-user studies of integrated museum and archive content resources like the OAC. The current status of this question with regard to MOAC is that the benefits of integrating access for the end-user outweigh the potential pitfalls. This opinion is informed by the ability to mitigate such pitfalls through such means as making descriptive method for each collection explicit instead of implicit for the end-user (an idea which is not yet in practice; but proposed). The EAD as a standard was developed to accommodate the range of archival descriptive practice, and this flexibility allows museum use as well. Initially, MOAC members worked with archivists toward writing one best practice guide for both archives and museums using the EAD for contribution to the OAC. However, it became clear that partners would be better served by having two different EAD implementation guides that ensured that core compatibility remained intact so that search and retrieval across finding aids in the OAC would remain consistent, but allowed museum and archive partners to meet their respective requirements more easily.

A telling detail of the project is the fact that when creating OAC best practice guidelines for both the museum and archival partners, with details on what EAD elements to use and how, it was found that the two community-specific guidelines overlapped by more than 80%. What this means is that, despite the differences between archival and museum descriptive practice, in an EAD implementation of such practice what museums and archives do is more the same than it is different. The remaining question, to be borne out in future end-user studies, is whether that particular 20% makes a significant difference for end-users or not.

Probably the most important finding of the MOAC project to date has been the answer to the question of cross-community integration. MOAC has found it to be very worthwhile and feasible to integrate collections access across the museum and archive communities. These communities are more similar than different in the grand schemes of the Internet and research methods and information, and they have much to gain by working together as well as by presenting researchers and the public with a unified face of humankind's shared cultural history.

The answer to the second MOAC research question, the appropriateness of the EAD for museums, is perhaps a bit more cautionary, but not for the reasons one might think. None of the museum partners in MOAC started, as a result of MOAC, to deploy EAD-encoded collections information on their own websites for instance. The one museum that had a previous EAD-based collections access system on their website, the Berkeley Art Museum, will be shifting in 2003 to a new access system, but will retain links to EAD collection guides in the OAC as another useful interface to those museum collections. Some museums, including the Berkeley Art Museum, however have been informed by the general idea of presenting collection as well as item level information. The reason why the EAD has not been picked up by museums has less to do with the archival origin of the standard than it does with the lack of inexpensive delivery and management tools for SGML and XML data. No XML based standard has enjoyed truly widespread deployment among the museum community yet. It appears that, for the time being, most small to mid-sized museums can start to think about using SGML and XML-based standards for the interchange of information, but it is not yet cost-effective to think about those formats as direct-to-end-user delivery formats.

MOAC has found that the strength of the EAD for the museum community is in allowing museums to share and integrate collections access with archives and libraries and to participate in large-scale content sharing projects. A secondary benefit of the EAD for museums is simply in getting museums to think about how to provide the full range of information they have on their collections, from collection to item level. The EAD however is not the end-all standard for museums; in fact no standard is. Each standard allows museums a certain set of functions but each such standard should be considered one tool in the museum's toolkit, to be brought out when the appropriate need arises.   For instance, the EAD will serve museums well as a format for sharing and exchanging collections information in collaboration with archives and libraries, or perhaps when it is beneficial to include collection-level information. The Dublin Core standard may serve museums well when they need to define a minimal record approach. To summarize; the finding of MOAC in this area is that museums need to become generally more capable of conversing in multiple standardized 'languages', and not come to believe that any one standard will solve the most important problems.

On the question of implementing EAD at the item level, MOAC is not alone in finding that it is entirely possible to provide very detailed item level description within the EAD. In order to test whether the EAD could accommodate the item level detail that museums would require, MOAC looked at the REACH project. REACH (Record Export for Access to Cultural Heritage) was an RLG and Getty-lead project to define a sort of "Dublin Core" minimal museum object record format that could be derived from typical existing museum collections data and still be useful for research access. MOAC took the REACH element set as the baseline for an item level museum record, and mapped that to the EAD. The EAD easily accommodated the level of item detail required by museums.

On the further question of testing how the EAD could accommodate linking from item level records to images and other media; MOAC (indeed OAC in general) found that the EAD is very robust with regards descriptive metadata, but other standards would need to be added to the mix to accommodate the complex technical metadata necessary to present media files. This finding is fairly clear, and has also been included in the recently released RLG EAD Best Practices document. Initially, MOAC partners needed the ability to contribute and present relatively simple images of museum artifacts, which typically included one thumbnail image and then a larger size or two of the same image. Quite soon however, some MOAC partners needed the ability to present more complex media representations of their items such as individual page images which together comprise the representation of an artist book, different views of sculpture, or multiple image representations of stereographs or Asian scroll paintings. In this area, MOAC was also testing how to provide very rich access to collections that might serve research needs beyond simple item identification. It was found inappropriate to use the EAD for this level of metadata as the EAD was made to contain data about a collection or collection item, but not to contain a lot of metadata about the media representations of that object.

In order to accommodate the one-to-many relationship between one item record and multiple images, MOAC needed an additional standard suited to "complex multimedia objects". The Berkeley Art Museum along with others adopted an early Perl script called "eBind" written by Alvin Pollock (which remained in use on the Berkeley Art Museum website until 2003), but MOAC soon moved to adopt the prototype MOA2 (Making of America 2) XML schema for presenting complex media objects. Among MOAC members, the Berkeley Art Museum implemented MOA2 by creating MOA2 XML files for 60 artist books. These files were, and are, hosted by the UC Berkeley Library where they are linked from the EAD finding aids in the OAC. Currently, many MOAC EAD collection guides still contain HTML links to images stored on museum servers, with a few containing links to MOA2 files. The OAC is now considering the adoption of the emerging METS (Metadata Encoding and Transmission Standard) for application to all media files related to OAC. This would enable not only better representation of complex multimedia objects, but would also allow partners to submit more detailed technical metadata about even simple images, and in turn allow the OAC to better inventory and manage the multimedia component of the OAC. Additionally, the Berkeley Art Museum has created TEI Lite XML documents containing transcriptions of text from the pages of the same 60 artists books. These TEI documents are linked from the MOA2 file at the level of a page. In this way a chain is created starting at the EAD document that describes the collection down to the level of the item. From there the EAD links to a MOA2 file which describes the media object in terms of it's structure (chapters, page numbers, location of page images) down to the level of an individual page, and from there the MOA2 file links to a TEI file which contains the text transcript of the page for purposes of legibility, screen readers for the blind, and text searching. Details of the relationships between the various standards MOAC ended up using will be detailed in the final MOAC report on the MOAC website in early 2003, but the summary of this finding is that MOAC has found that in order to provide access in an EAD-based system to museum materials, which are usually image-heavy, one should include other standards, such as METS, SMIL, or MPEG for access to the related media files.

On the question of semi-decentralized content storage (mainly image files stored on local museum servers), MOAC found that this works on a basic level, and could work in a more robust manner if certain procedures are put into place. In fact, most of these procedures are not yet in place with MOAC, but the project is heading there. The first observation of this approach was that online presentation did not suffer as a result of slower server response speeds from local servers holding images in comparison to the central server hosting the EAD finding aids. MOAC museum partners appreciated having more control over their image content, and in one case the museum had to quickly take an image offline due to copyright issues; and this was easily done as it was on a local server. However, such local control comes with a price, and has pointed out the need for additional policies. For instance, the image in the example above was taken offline before the EAD collection guide on the central server was updated, so the text appeared to users with a broken image link for a time being. Clearly, MOAC will have to establish procedures for quick content updates to either text or images. The benefit of local control of images is attained at the expense of a certain amount of certainty from the central system point of view. MOAC has determined that in order to allow decentralized hosting of content, partners should be required to meet certain minimum infrastructural requirements, and should furthermore agree to a policy document which stipulates how much notice is necessary when images are moved or taken offline, how much downtime is allowed for local server maintenance and crashes, and what level of network speed and technical support the local image server will be required to have to ensure that the central resource can provide stable service. It could be the case that such a policy document could be expanded via the professional community into a semi-standardized way to create an institutional readiness profile' for hosting shared content which could benefit multiple projects.

On the surface, such an experiment might seem needlessly complicated. Why not just put copies of all the images on the central server? Indeed MOAC is investigating this option to enable such functions as backing up image files or providing sophisticated images services such as the ability to zoom in via Mr. Sid or some other tool. This experiment in decentralization is worthy however, because it takes a practical step toward a possible scenario that has been imagined for years in which cultural heritage institutions can store much of their content locally, for maximum control, flexibility, and timely accuracy and allow this content to be queried, in real time, by various portals and tools elsewhere. This scenario has been explored by efforts from the development of the Z39.50 protocol to more recent OAI harvesting testbeds. MOAC in collaboration with the California Digital Library is in fact considering a harvesting component for a project that would create an image repository. For now, MOAC sees the value of centralization and redundant content, but remains interested in practical steps that explore the implications of decentralized or hybrid content hosting models.

To conclude the report on the initial research questions, it should be noted that perhaps the second most important finding of MOAC, after the feasibility of museum and archive collaboration, is the finding that it is realistic and practical for mid to small museums to utilize even complex standards such as EAD, and to participate in large-scale, standards-based consortia projects. MOAC included partner museums with total staff of less than 10, and is confident that the procedures and workflows that have developed during this testbed project will be highly adaptable to all sizes and types of museums. It should certainly not be understated how much time and effort are still required to undertake participation in large-scale collaborations, the effort is Herculean especially when an institution is just starting out, but there are several key elements that MOAC has found can make the effort feasible, cost-effective, and worth the end result.

The key elements to have in place that help ensure successful collaboration are, firstly, professional community. One of the benefits of a regionally based collaboration is that members are within easy visit or phone calls to each other, and MOAC members have retained an open attitude to helping each other with advice via the MOAC email list or a visit. Part of the success of this community is of course its current small size, but larger communities have proven successful online (sometimes more successful) when a critical mass of ready help is reached, or when larger groups can be effectively split into smaller support groups clustered around specific topics or problems. This community needs to be a focus and not just a by-product of the relevant project.

The next element is to have realistic expectations. During the entire ongoing process of writing the specs and guidelines for contribution to MOAC, members kept a firm eye on what was practical and feasible. For instance, MOAC specs were written to utilize only existing museum data, to allow each museum to retain individual collection management systems, and workflows were developed to dovetail with existing practices. The price of this restraint sometimes is certain sophisticated end-user functionality, but it was deemed more important to start with something and build out than to put obstacles in our path early on.

The last element that needs to be in place is tools. Museums, especially small museums, can only get so far with realistic specs and good advice; they need tools in the form of software that can aid their processes. This last element became so important to the project that it deserves expanded discussion below, but here is a quick summary. The use of standards greatly enhances a project's ability to scale up, offer flexibility, and to achieve longevity. It has been held by many that such standards however are developed by large research institutions and are often too complex to be adapted by smaller, resource or expertise poor institutions. This viewpoint is actually largely true, but is not fatally true. While the cultural heritage community does indeed need standards which meet the above criteria of creating realistic expectations of the least of us, it is possible with the right community, a practical interpretation of the standard, and the right tools, for even small institutions to participate in building our shared cultural heritage resources.

What MOAC Has Learned That We Did Not Expect To Learn

In addition to finding answers to many of the questions on the initial research agenda, MOAC has found additional unanticipated answers to unexpected problems. Some of these are the most exciting, and most formative developments of the project.

Foremost among these was that a new emphasis developed on lowering the cost of participation. Initially the focus of MOAC members was very much toward the center and toward what technical and metadata specs were needed to create a useful, integrated content repository. Into the project, MOAC had established some of those target specs for partners to aim for, and practical production issues forced MOAC members to turn their attention back toward the individual institution and how to enable that institution to achieve those target specs. The most significant development toward that end was the creation of a tool, developed at the Berkeley Art Museum, and called the Digital Asset Management Database (or DAMD for short, with all due love-to-hate sentiment). This tool, implemented as a FileMaker database allowed the Berkeley Art Museum to import collections records from the museum collection management system; to then add technical metadata generated during the imaging workflow; to marry the descriptive records with the images; and most importantly to automatically export the necessary data as fully-formed EAD SGML collections guides which complied with the MOAC best practices guide, and to export finished MOA2 and TEI XML documents when desired. This tool was shared with other members of MOAC, and several have now had experience using the tool to generate the standardized files to send to the central server by way of contribution to MOAC. A couple of other museums and archives developed similar tools that they continue to use as well. DAMD will be covered in more detail in the final MOAC report on the MOAC website, but one of the key things DAMD and these tools did was to remove from the contributing institution the necessity for a certain level of expertise with the multiple, complex standards; EAD, MOA2, and TEI. This is not to say that each museum did not need staff that knew how to input their data in such a way that it would make sense as output EAD, but certainly those staff no longer had to know how to actually encode and create EAD files. This approach does not reduce the importance of the standard to the project, but does reduce the cost of applying the standard.

DAMD has gone through several iterations, and continues to be developed. The Berkeley Art Museum is currently looking into ways in which it can overcome intellectual property and support issues, and share the DAMD tool with the cultural heritage community in an open-solution manner. MOAC has already come to the conclusion that, just as with standards, no one tool will solve every institution's needs. Different institutions have different platform support and different workflow needs that have already required the development of a couple of tools. The important thing to learn from this is that it is important to tackle the problem of how to lower the cost of participation if museums and archives are ever to achieve broad-based participation in scaleable collaboration projects.

Other unexpected findings include the benefit of a regionally based project. Initially, it would seem counter to the nature of the Internet that any such collaboration need be geographically bounded, and ultimately projects like this need not be. But regionally based projects benefit in untold ways from the ability of partners to communicate in the same time zone, to make visits easily to tour each other's facilities and get advice, and to simply run into each other serendipitously. The other benefit of a regional project providing intellectual access to real, physical objects is that researchers and others in the same region who use the content resource know that whatever collections they discover in the system are within a few hours drive from where they are. Physical access is not of course automatic, and should not be overstated as a complete rationale for making a project regionally based, but it is a side benefit. The relationship between intellectual access to collections information and research access to the physical objects deserves much more research. It is clear that online access clarifies the need for clear physical access policies to avoid over-use and wear. Having said that, some physical use is desired and can be enhanced by online access. For instance the Berkeley Art Museum has seen more requests for visits to the Theresa Cha Conceptual Art Archive, which went online early in MOAC, and those requests are more precise because researchers can search online before they visit to determine the exact objects they need to see - thus reducing the needless wear on the collection. Faculty at UC Berkeley has also begun to use the Berkeley Art Museum's online collections in their instruction and to prepare for a visit to the museum and the physical collection.

Some MOAC partners have found that commercial interest, and revenue opportunities, have increased as a result of their content (mainly images so far) residing in the larger context of the whole OAC where are available beyond the museum's own website. This increase is not significant enough to be the main reason for participating in a project like MOAC (in part because it has not been developed in any way), but it is encouraging to see that in some small way MOAC could support open public access and yet still receive some economic return, however small.

As the reader will note in the next section, some of these unanticipated results are helping shape the future of MOAC, and as MOAC continues to evolve, it will be important to continue to listen for answers to questions we forgot to ask.

Digital Content Distribution and Digital Content Aggregation

As mentioned earlier, one of the most important recommendations from MOAC is that the cultural heritage community needs to continue research, testbeds, and production in the area of aggregating content for the end-user, and we need to add a new focus on the ability of individual institutions to prepare and distribute content. Many cultural heritage community projects to date have approached the problem of integrating content from the centralized angle of federating the content; developing standards, specs, ingest processes, and portal tools that allow digital content to be brought together and delivered to the end-user. Not as many projects, nor as much attention overall, has been devoted to the other side of that equation; tools, guidelines, practical specs, and models that enable the individual museum or archive to distribute content to more than one portal easily and cost-effectively. MOAC also initially concentrated on how to federate content and how individual museums may contribute to one portal, the OAC. More recently, in partnership with the CDL, MOAC has begun to explore how to distribute content more easily. Other projects, particularly those developing harvester tools, are also exploring distribution, but there is still not enough research from this angle, and too few workable models for institutions.

MOAC has seen the need for new institution-centric models for distributing content that are as robust, scaleable, and well-conceived as our portal-centric models for aggregating content. Whereas portal-driven projects are usually oriented around several contributors and one portal, institution-centric projects need to invert the relationship, orienting toward one institution's relationship around several portals. It almost goes without saying that the cultural heritage community needs an equal amount of energy on both types of projects since without capable institutions, portals will not grow, but without portals, institutions will not have a place to deliver their integrated content. Given this condition, both types of projects should continue to combine partners; content-producers and distributors with content-aggregators and deliverers.

Future of MOAC

The IMLS-funded testbed phase of MOAC draws to a close on December 31, 2002. The testbed produced 20 collections from 10 museums, including about 50,000 images, as a means to test the production and integration issues outlined earlier. These resources are publicly available online at the MOAC website, and will remain part of the growing museum content in the OAC. MOAC will release a detailed project report, including this narrative report, on the MOAC website in early 2003.

MOAC members and the California Digital Library agree that the testbed has proved the desirability and feasibility of museums' participation in the OAC, and all sides have agreed to move the collaboration from testbed phase to ongoing production. One of the first moves of the project from testbed to program is to open up participation from the initial 10 to any museum in California. The first new museum to join was the San Francisco Museum of Modern Art, which brought David Sturtevant, Dana Mitroff, Susie Wise, Marla Misunas, and Tim Sevonius onto the MOAC Steering Committee. In addition to continuing the original project, MOAC has made several other major expansions and changes.

The first of those changes is that MOAC is no longer one project. During the course of the original MOAC testbed project, new opportunities for collaboration have opened up, and specific areas of inquiry from within the original MOAC project have by necessity spawned whole new projects. The first of this new suite of projects is the continuation of the original MOAC project to encourage and enable museums to participate in the OAC by producing and contributing digital content (this effort has been dubbed "MOAC Classic"). The most closely related of the new projects is a new IMLS funded effort to conduct end-user evaluation and surveys on the MOAC content resource; to establish how the resource is currently being used and what improvements could be made to the content or interface in the future. This project is referred to as "MOAC II", and includes a sub-set of the original group of museums along with the CDL. In another project, the newest University of California campus, UC Merced, Library will be working with the Ruth and Sherman Lee Institute for Japanese Art to digitize the art collection, with advice from the Berkeley Art Museum and CDL, and contributing that content to MOAC. In another project, the Berkeley Art Museum and UC Berkeley Library are using content they have contributed to the OAC and working with the UC Berkeley Interactive University to create prototype tools to enable teachers and faculty to more effectively use that content in teaching. In another project, CDL is working with MOAC members, Luna Imaging, and others to create a prototype image repository and service that aggregates images and metadata from MOAC, AMICO, and UC slide library projects among others. Last, the Berkeley Art Museum and others are forming a project to explore the further development and means to broadly share the DAMD database, developed and used by MOAC, as a museum community tool for managing media metadata and more importantly converting that data into various standardized formats to enable smaller museums to contribute to a number of content portals, including RLG's Cultural Materials Initiative, AMICO, and, naturally, OAC.

In recognition that MOAC is expanding beyond one project, including participation in more than one CDL initiative, MOAC members agreed to change the full name of the project from "Museums and the Online Archive of California" to another name which retains the identity and spirit by keeping the M-O-A-C acronym, but unfolds to reveal an updated and more accurate sense of scope. "Museum Online Access California" has been proposed, and is pending further discussion, changes, and approval. MOAC is still MOAC, but the focus has shifted slightly from a project involving museums and archives, to a group of museums and archives collaborating on several projects. MOAC members unanimously agreed that we had achieved something successful, and did not want to lose momentum or focus, and that the goal was simply to update the organizational model slightly to accommodate more than one project, but not to significantly change the business model or spirit of the collaboration. The group made it clear that we do not want to spin off a new, separate, non-profit entity that would necessarily consume administrative energy and resources, but rather to keep the focus on the group of institutions working together primarily with the OAC, and on new projects like those mentioned earlier.

The new organizational model then is still a basic, non-profit consortium consisting of museums and archives, with each member naming a delegate to the MOAC Steering Committee. Richard Rinehart will chair the consortium for the first year, at which point MOAC members will revisit and choose a new rotating group chair. Most of the detailed working organization will exist within the projects themselves, and not be centralized. In these ways MOAC will explore, in an organizational as well as technical way, an institution-centric and de-centralized collaboration model. The majority of MOAC members are museums, but in continuing to work on "MOAC Classic", and with the CDL, Bancroft Library and potentially other archives as partners, MOAC will continue to make a unique contribution in the area of cross-community collaboration.