 This presentation explores the choices digital programs face as they add functionality to an expanding corpus of content and services. It focuses on archives and academic library special collections with metadata as the particular lens for this analysis. The scope, though, includes all aspects of digital programs and the concept of interoperability frames the discussion. We will look at several ways to conceive of it. Perfecting interoperability is a bit of a holy grail for digital and special collections since the content and culture of archives gravitate toward bespoke solutions. Unique content, the purpose of archives, can seem at odds with uniformity. But uniformity is what can bring multiple elements of digital programs for special collections into harmony. My name is Michael Herrick and I'm a cataloging and metadata librarian at San Francisco State University. I started here in September 2022 about a month before we began a three month process to migrate from content DM to Cortex, the Dams and Content Management System marketed by Adam Matthew, now AM Digital. Adam Matthew developed Cortex to produce and deliver curated digital collections of primary source content. AM Digital now offers the software as a service to enable individual institutions both to manage their digital assets and to arrange content into multiple types of collections. This front end content management part of Cortex is fairly customizable and thus speaks to an archivist's affinity for unique. Before coming to San Francisco State, I worked as a manager focusing on metadata and description at the Hoover Institution Library and Archives on Stanford's campus. My time at Hoover was dominated by participation in the planning for and launching of its next phase and a progression toward delivery of its fast and varied archival collections over the web. The front end of their Digital Collections 2 beta site, first put on the web in rough form in 2021, is truly but the skin of a major effort to rethink digital delivery for archival researchers. As the beta website currently states, Hoover is developing an ambitious new digitization program with an emphasis on digitizing whole collections to uniform standards. Hoover's collections serve an international body of advanced researchers. The library and archives are 100% privately funded with support coming from a relatively small body of donors. San Francisco State is a publicly funded university with a commitment to quality teaching and broad access to undergraduate and graduate education. The archival and unique collection materials of both institutions sit within larger networks of ownership and digital opportunity. Situated on Stanford's campus, Hoover has access to its digital repository built within the San Vera framework. San Francisco State is part of the 23 campus California State University, the largest public university system in the United States. The CSU has also adopted San Vera tools recently and San Francisco State is building its institutional repository for master's theses and other content in this direction. For digital archives though, neither Hoover nor San Francisco State chose to deposit assets to their respective broader institutional repositories during their most recent rounds of decision making about digital. A final note about the obvious serrations I'm making. While I was intensely involved in discussions about strategic decisions for digital collections too at Hoover between 2018 and 2021, I ended my involvement with that program in 2022. Conversely, I came to San Francisco State to help clean, wrangle, and ingest metadata in the recent migration to Cortex, but I had no involvement with the strategic decisions made here in the period 2019 to 2022. My reflections therefore are based on internal observations of process, the weighting of factors for decisions, and the consequences of these decisions, but I'm not really overwetted to either process at this point. These observations are mine and do not necessarily reflect programmatic statements of either institution. I've created this dual project briefing to help all special collections librarians and archivists to consider the challenges and factors at their home institutions as we all move ever more content online in creative and impactful ways. Anyone who has worked with digital migrations, implementations, or really digital content in any form in the back end knows that the glossy skins with which researchers interact on the web belie the complexity of decision making and trade-offs that go into a full digitization workflow and a digital publishing program. Many hard decisions must be made about interrelated components and resource allocation. When drilling down into the details of each component or node of a total program, one can spawn telescoping or metastasizing pathways of opportunity or need. The work simply to scope everything can absorb vast amounts of time and energy. In the next 10 to 15 minutes I will not cover everything about these two projects or implementations, but in migrations and implementations one needs to consider the fact that long term everything must interrelate or become interoperable on some level. Therefore, I'll spend a few minutes talking about interoperability generally before moving on to a brief description of the two projects in a contrast to compare format. Finally, I'll provide some reflections on metadata and long term strategies we should be considering as we balance two things. First, the established needs and examples of metadata surfacing in single aggregator websites like the Digital Public Library of America and Europe. Second, the emerging needs of linked data applications which offer the promise to build niche aggregators on the fly and via machine-based methods. When we think about interoperability in digital archives, both the International Image Interoperability Framework, TIPLIF, and the Open Archives Initiative Protocol for Metadata Harvesting, OAIPMH spring to mind. These are two enormously successful contributions to interoperability for finding and using digital content on the web. I strongly endorse their use, but the way we'll talk about interoperability here aims both smaller and bigger. Locally, I am talking about the interoperability of all the components of digital workflows. No one implementation of a system, software tool, adoption of a schema or workflow exists in isolation. Declaring adherence to principles like FAIR or implementing a TIPLIF viewer are important, but without a lot of effort devoted to the interoperability of local processes, digital production and dissemination quickly grind to a crawl or total halt. Every digital program aspires on some level to balancing all seven elements on the screen. Almost no program can make progress with all seven simultaneously at an even pace. Therefore, an implementation of a new system or element can be thought of as part of a dance over time. Less successful programs are the ones which just keep trading in one-to-one dance partners to keep up surface level confidence and an image of competence. It is important that they keep dancing and they will never be left without a partner or at least one individual connection to the total dance. Programs aimed at long-term capacity building and success see the full dance floor as an opportunity for growth. This approach includes all elements or all dancers. Long-term successful programs move forward roughly in sync, keeping all parts dancing to the same rhythm. This is interoperability over time for realistic digital production. At the end of this briefing, I'll reflect on the bigger aspect of interoperability, which I see as more of a metadata-specific topic. Catherine Ruddick gave a presentation in this pre-recorded project briefing series in April 2022 that provides an excellent example of implementing an all-in-one solution. The image on the right side of the slide is from her presentation, which I recommend you watch in conjunction with this one. The data or digital curation lifecycle derives from images on the vendor's website, Orange Logic. This type of dam's implementation approximates, as much as possible, all-in-one interoperability. Neither San Francisco State nor Hoover elected to take an all-in-one approach. Each local decision-making process follows its own logic based on its stakeholders and needs. But the processes at these two institutions shared a common factor in that that desire to solve one of the seven elements on the preceding slide weighed much more heavily than the other six. So an all-in-one dam's implementation was never really considered. That is not necessarily negative. All-in-one solutions have their strengths but also drawbacks by limiting versatility over time in doing the dance. In a couple of minutes, I will show a slide that captures the total weighted picture of the prioritization process at these two institutions. In terms of local interoperability, I will note that the AM Digital Cortex Solution is, in a sense, an all-in-one dam's directly linked to content management and publishing. But when born digital collection, collecting, digital preservation, and other elements of the total digital life cycle are considered, it is clear that Cortex, in its current form, is heavily skewed toward publishing and curating collections of materials for front-end defectiveness. Interestingly, Ruddock notes that the Cortex dam's, that's C-O-R-T-E-X at the University of Calgary, is linked to LibNova's LibSafe product for long-term digital preservation needs. The LibSafe digital preservation platform forms the core of Hoover's solution in that it is sandwiched between Archive Space and LibNova's discovery layer called Open Access. Hoover and the University of Calgary share the choice of LibSafe for digital preservation. The similarity of roots taken ends there. The methodology for decision-making at Hoover shared a lot, though, with that at San Francisco State. Even though the dominant priorities for each fill in opposite ends of the preservation and access continuum, each institution followed a similar path in how they researched and deliberated about a new implementation. Waiting priorities is not the same thing as prioritized waiting. Waiting priorities assumes a sequential process in which all priorities are scoped and established. Then research and a decision-making process occurs to prioritize and to allocate resources within a given time frame. Prioritized waiting, which I think is relatively common and pretty much happened at Hoover and San Francisco State, means that one or two elements on the total dance floor of digital interoperability receive preferential consideration at the start of a new phase of planning and implementation. Phasing the upgrading of elements on the total dance floor of digital interoperability is inevitable. Almost no institution is starting from scratch. Complete digital programs, which almost all archives will aim for given their mandate both to preserve and provide access, are thus required to phase change to the various elements. This is the only way to evolve ever more complex and effective digital services. Over the past decade, Hoover and San Francisco State have done exactly that. Nearly a decade ago, both had already brought digital asset projects, begun in the idiosyncratic manner, which all programs developed in the 1990s and early 2000s, into a small number of management systems. At San Francisco State, this was content DM for everything but the Bay Area Television Archive, which was providing access to digitized content through a locally developed platform called DIVA. Almost 10 years ago, Hoover also brought all but a few key groups of assets into gallery systems TMS. A main common driver for both institutions to seek replacement solutions in the past years involved the workflows for and eventual surfacing of OCR and audio transcriptions to enhance discovery and access. But their motivations for this derive from their differing priorities. The Bay Area Television Archives established a relationship with Cortex in 2020, two years prior to the general migration, largely for accessibility issues. Getting content fully transcribed, a feature which Cortex was building directly into their dams, assists San Francisco State's mission to provide equal access to all communities and to foreground accessibility to the content which treats the history of the whole Bay Area in the 60s, 70s, and 80s. At Hoover, full text searchability, whenever feasible, has become a key element in their ambitious goal to digitize entire archival collections in a uniform way. With human-produced metadata being a prohibitive cost element to mass digitization efforts, Hoover developed a strategy by which digitization could happen based on the slim metadata contained in the EAD of finding aids. This allows entire collections to be made available to researchers soon after digitization. The API-based Archive Space LibSafe connections will give researchers access through the finding aids almost immediately. With a focus on first getting things preserved and then getting as much full text access as possible built out as part of ingest, future stages of Hoover's original plan envisioned providing large textual datasets to researchers. It was also envisioned that machine-based entity enhancement in the total-meditated corpus could eventually build more and more discovery pathways. Both Hoover and San Francisco State have long-term visions while accepting short-term trade-offs. That San Francisco did not solve digital preservation up front does not mean that they are not as equally committed to this goal long-term as Hoover. That Hoover, in the beginning, sacrificed content management functionality beyond basic research or needs in order to build a fundamentally solid core system for digital preservation does not mean that they aren't equally committed to building strong content management solutions for niche publishing and teaching in order to highlight small areas of their total digitized corpus. In the last part of this briefing I will share a few reflections about interoperability as it relates specifically to metadata. The smaller focus of interoperability, which I've already presented, helps to keep all pieces of a local program dancing roughly to the same rhythm over time. Metadata is but one piece of this dance. As libraries, archives, and museums, or LAMS, edge ever closer to an environment in which linked data tools and platforms drive interoperability, I think a shift in perspective about metadata for digital objects within archives could happen. Near the start of the Digital Collections 2 planning phase, a couple members of Hoover's very small planning team were invited to participate in a two-day session at Stanford, which brought together people from libraries, archives, and museums to brainstorm further growth of already highly successful IIIF development. Within this two-day LAM melting pot, there was a strong drive to force consensus around uniformity of metadata elements in IIIF manifests. The museum community seemed convinced that it should be possible to agree on title and a few other elements to achieve metadata interoperability across communities. Librarians seemed convinced that linked data development and IIIF development would and should proceed with awareness of each other's domain without rigid forced integration. Archivists seemed somewhat perplexed by the whole conversation. Build out of extensive descriptive metadata for archival objects taken out of the archival context does happen in digital archives, but the original order of creation, storage, and presentation provides the fundamental framework of archival description. Traditional archival description in aggregators is both not effective and can become astronomically expensive. To make a description of archival objects aggregator friendly is like completing all the work which a creator and a publishing house need to do when they prepare a book or article for publication. Once all the descriptive elements have been created, they then need to be formatted into a structured form with multiple access points like cataloging and publication or CIP statements. When the original creator of an object does or oversees this, the resulting description tends to be accurate. An archival processor or digital program metadata producer works at a certain number of degrees of separation from the original context of creation. Changes in cultural, historical, and linguistic context result in blurred accuracy. By assigning an archival resource key or arc to every object and building a metadata application profile at Hoover which is focused on the eventual generation of RDF triples, an underlying framework has been established to make the atomization of interoperability for metadata creation and use possible. OAIPMH and the main digital object aggregators of the past several decades necessitate a level of description which is simply unrealistic when working with mass digitization of archival content. People are successfully doing it, but scaling it for mass digitization of archives is not economically feasible. The emerging potential to provide a wiki-based type environment for creating collaborative knowledge bases could greatly expand the discovery potential for digitized archival content. For Hoover, this is basically essential since its collection mission has been focused on bringing in content from all over the world in a huge number of languages. Quality discovery pathways to individual items will need to be aided by both machine applications and worldwide collaborative effort. The same approach would work equally well at San Francisco State. This would marshal the collective knowledge and participation of the entire San Francisco Bay Area instead of relying on a small number of staff on the San Francisco State campus. Both San Francisco State and Hoover are still very much in the soft launches or beta phases of their current implementations. I encourage you to check back at these sites in the coming months and years to see how they each build functionality and at what pace. The main tensions they experienced in planning for this phase involved how to balance resource expenditure along the preservation and access continuum. My discussion of the linked data potential for digital projects and archives has been brief, but I encourage you to think about the possibilities for deep linking and discovery across archival collections and repositories. Content can continue to be represented in major aggregators, but some of the most fruitful connections which will lead to the powerful interoperability will I think happen when RDF statements begin to create direct connections between related content to positive archives throughout the world. Thank you for listening.