Discovery Feb 2012

Abstract/Agenda: 
  • Recap of Winter Meeting (if needed)
  • Status of new DCPs
    • DCP-4: Use of xlink attributes in Atom <link> tags
    • DCP-5: Use of OpenSearch <Query> tags for valid parameter values
    • DCP-6: Replace overloaded time:start and time:end tags with dc:date
    • DCP-7: define error handling best practices
Notes: 

Recap of Winter Meeting (if needed)

DCP-4: Use of xlink attributes in Atom <link> tags

  • Current Version: DCP-4: OPeNDAP Links in the Atom <link/> element
  • Potentially use "role" and "arcrole" from xlink to capture all the semantics of DCP-3.
  • Should we change DCP-4 to the current OPeNDAP suggestion, or should there be a separate DCP for the more general case of DCP-4?
  • Brian suggests a separate DCP for the general case (e.g., using "arcrole"), this will be coming in DCP-5
    • Basically, covering the case where service casts refer to collections, or collection casts referring to services
  • Returning a collection cast is a different use case from returning OPeNDAP response
  • Pedro suggests you don't need to use something like arcrole, IANA has registered mime types and rels that should cover everything
  • Summary -> DCP-4 will use xlink role, DCP-5 will use xlink role and arcrole

DCP-5: Use of OpenSearch <Query> tags for valid parameter values

  • Likely renamed to DCP-8
  • Li is interested in this, will be in contact with Ruth Duerr and Discovery mailing list for further information

DCP-6: Adopt Dublin Core Date Specification in Atom Response

  • Current version: DCP-6: Adopt Dublin Core Date Specification in Atom Response
  • More aligned with OGC use of OpenSearch
  • dc:date refers to "point or period of time for an event associated with a resource"
  • should we use the convention that the parameter and XML element should match?
  • dc:date is more standard than the OpenSearch parameters
  • time:start and time:end were only standardized as "queryable"
  • which does ESRI prefer or suggest based on client development?
    • ESRI would align more with the dc approach (since there is already a dc:date XML element)
    • Not a big deal either way, but prefer things that may become standardized
  • another option (considered by OGC) is using gml, but this is more complex
  • Brian seeking connection between parameters and response elements in OSDD; however, OSDD does not support this
    • Only potential mechanism is "rel" and "type"
  • Eric also suggests using the "rel" list capabilities of OpenSearch to tell clients of this "convention"
  • The response may not actually echo all the values for search parameters
  • Another avenue is to use the Query tag is someway
  • Could use the <atom:Content> tags to put whatever data you want

DCP-7: define error handling best practices

  • There is no standard for how errors will be handled
  • Formally adopt HTTP 1.1 status codes
  • HTTP 400s cover client side, 500s cover server side
  • Lots of different options for responses, Atom response containing error, HTTP error code, etc.

Originally posted on the wiki: http://wiki.esipfed.org/index.php/Discovery_Telecon_2012-02-14

Citation:
Lynnes, C.; Hua, H.; Discovery Feb 2012; Telecon Minutes. ESIP Commons , April 2012