NOTE: The galeon mailing list is no longer active. The list archives are made available for historical reasons.
Wenli, This issue of "granularity" or heirarchies or collections or groupings of datasets that are alike in some way was one of the issues confronted early in the THREDDS project. As a result, I believe we have an approach that works reasonably well in the THREDDS Data Server package. The issue continues to arise in most discussions of data and metadata collections and services. In fact it was one of the issues discussed at the 3rd Metadata Interoperability Conference I attended last week. It will be important to confront it in the context of OGC and ISO standards. The disadvantage of doing it in the WCS context is that one can envision collections that might include Coverages, Features, and Sensor Observations. For example a collection of all the data related to a specific event such as a severe storm, a flood, a hurricane, and so forth. One can create THREDDS catlogs for such "case studies." But it would be good to eventually have a standards-based interface for such collections. Perhaps the OGC CSW is not well suited to this sort of use at present. If so, it may be useful to consider suggesting augmentations to CSW. I believe there is a big advantage in that we already have a working system. I plan to send a copy of this to the THREDDS and GALEON groups as well as to John Helly who convened the Interoperability Workshop last week. Thanks for your careful description of the issues in terms of THREDDS catalogs and OGC CSW.. -- Ben On 11/15/06, Wenli Yang <yang@xxxxxxxxxxxxxxxxxxxxxxx> wrote:
Ben, THREDDS deals with service/data hierarchy nicely. However, I think that CSW does not provide guidance/standard on how hierarchical service/data should be presented. When mapping a THREDDS catalog into our CSW, we can track and record the hierarchical relationships among data/catalogReferences and among different levels of catalog references in our database. We haven't fully investigated how such relationships can be presented to a CSW client (or how a CSW client can request such relationships). This is certainly a very useful piece of information and deserve further discussion. I have not carefully read the WCS hierarchical description part which was primarily provided by Luc. I think that the primary intention of using hierarchical description in WCS capability was not to let a client actually retrieve this the hierarchy information but was to reduce the duplication of metadata in, and thus the size of, the capabilities document. Initially, it was hoped that the hierarchical information would allow a client to retrieve a collection of data sets (coverages) from a higher node in the hierarchy but it was decided that this would not be specified. Of course, a specific server implementation can still provide such capability by declaring a collection of coverages as one single virtual coverage. For example, a THREDDS service reference containing a time series collection of data sets (individual coverages) for a specific location can be declared as one coverage with a time span covering all the data sets. In addition, each of the data sets in the collection can, if needed, also be separately declared as a coverage with time range being at a point time (or a smaller time range as compared to that of the collection). Wenli
galeon archives: