On September 15th and 16th, 1997 the Second IEEE Metadata Conference was held at the National Oceanic and Atmospheric Administration (NOAA) complex in Silver Spring, Maryland. The main objectives of this conference series are to provide a forum to address metadata issues faced by various communities, promote the interchange of ideas on common technologies and standards related to metadata, and facilitate the development and usage of metadata. Metadata'97 met these objectives, drawing about 280 registered attendees from ten different countries and over one hundred different institutions. The audience included scientists, information technology specialists, and librarians from communities as widespread as finance, climatology, and mass storage. The technical program included two keynote addresses, two panel presentations, as well as twenty-three papers and thirteen posters selected from over one hundred abstracts. We provide highlights of the conference below. For more details, the proceedings are available electronically from the conference homepage at: http://www.llnl.gov/liv_comp/metadata/md97.html.The keynote addresses were "An Architecture for Metadata: The Dublin Core, and why you don't have to like it" by Stuart Weibel, OCLC, and "The Microsoft Repository" by Philip Bernstein, Microsoft.Weibel's talk described the Dublin core and the Warwick framework - a series of workshops whose output has been a core set of metadata elements common to data from most domains, along with a "container" based mechanism for plugging in larger domain-specific sets of metadata, like the FGDC's standard for geospatial metadata. These efforts represent two of the defining works in this community. Weibel touched on the history of this effort and described his belief that standards such as RDF (resource description framework) for the WWW coming from organizations like the World Wide Web Consortium (W3C) will have a major influence on the metadata community in the near future.Bernstein's talk covered the Microsoft Object Repository, which also has the potential for a large impact on what metadata gets stored and how they are managed. Bernstein describes the repository as "a place to persist COM objects" (component object model), and as more than just a object-oriented database. The features of a true repository are 1) objects and properties, 2) rich relational semantics, 3) extensibility, and 4) versioning. Repositories are used to help tools interoperate by storing predefined "information models". The information models are the metadata used to describe the underlying COM objects in a standard way such that the objects can be shared across tool boundaries. The main consumers of this type of technology are tool vendors.