Servicemodel Ad Hoc Meeting held at Oracle, June 4, 5, 2001
Updated: June 13, 2001 17:56 - Page Maintained by Allan Doyle

Present - June 4 - about 60 people

Present - June 5 - about 25 people

Agenda/Minutes

Begin at 1300 on June 4.

Discussion:

Terminology has to be revisited. We have an opportunity to influence the terminology in 19119. ISO editing committee will be here on June 7, 8.

Goals for this meeting - i.e. what do we want to get out of the rest of this week's meetings of this group?

• How is OGC going to fit in with the rest of the world.
• Consistent terms
  • We need to use more discipline e.g. service vs. server
    service chain (Action: WSFL URL to be supplied by Lou - WSFL is available here)
  • 19119 terms need work… by June 6!!!
• Heterogeneous world
• Geospatial Services in the Web Services world
• Use case for service discovery & invocation
  • UDDI does not satisfy what we need
• They are at the same state we're at.
• Decision or path of A, B or B' (action: Allan to add a third diagram here - don't need a diagram, B' was a wording change. See "vote-definitions") or other
• Type hierarchies
  • WSDL root port type
• What is our interaction model?
  • Binding (we should explicitly say bad binding is bad).
  • Chaining, flow engine
  • Stateless, stateful
    • Sessions may be ok, state is ok, but don't require persistent connections to maintain the state/session, browser cookie style mechanisms are "out of band" and are to be avoided (or at least not be required)
      Bottom up vs. top down "meeting in the middle"
• Is WSDL high enough up?
  • Only gives single-service view
  • C.f. ICE testbed view
• Meeting (or top down vs. bottom up) is a natural consequence of going about this intelligently
  

Time frame

• What's the time frame for doing all this?
  • September '01
  • WMT-3 RFQ issue date - End of June '01 ??
  • Really, we need answers by the end of this week.

   

Arliss' Topic 12 Revisions - 01-045

[Note: Arliss presented this here because the Architcture SIG is not meeting during this TC week.]

Discussion:

• Clarification on conformance - should an AS have a conformance clause in it at all?
• Should 19119 be a part of Topic 12 as opposed to being all of T-12?
• Doc SC wants to have appendices to ISO docs detailing diffs.
• Last time we voted to accept on instance of 19119 not all future instances. Precedent is that we voted on geometry each time it was changed by ISO. We should document what the changes are. (does this cause copyright issues?)
• OGC should not get bogged down in what the ISO is doing but should also not ignore ISO. TOCG to take this up.
• 19119 is more dangerous. Adopting external spec has to be carefully thought out. What about the IP's influence? We should still be able to let the IP influence the SP.
• 19119 editing period this Thursday/Friday. Most comments are from OGC. In this case the process is not harmful to OGC.
• Editor told us that the OGC comments were to be collected before a previous deadline, not realtime comments from this meeting.
• Don't believe that OGC can depend on having another organization change our documents, but we also should make use of other organization's work.
• OGC Topic 12 change process should be the OGC AS process, not the ISO process. I.e. don't use ISO as an editor of OGC specs.
• NASSL spec influenced some comments to 19119, was fed into comments. Relationship between OGC, W3C, and ISO is critical to some gov't level bodies (e.g. Australia)

[Note - there was almost a motion to adopt 01-045 with minor modifications however an objection was raised by John Herring that since we were not meeting as a SIG or WG, we had no basis for bringing forward a motion to the TC. The motion was deferred to the ICS SIG meeting on 8:30 June 5 (where it passed unanimously with no discussion)]

WSDL + GML - Ron's slides… (Action: Allan to get slides and reference them here -- here they are PPT, PDF)

• Ron's proposals:
  • GML as syntax for service metadata
  • GML as syntax for messages

WSDL discussion - lots of sentiment in favor of moving this way

Adjourn at 1500 on June 4.

Tuesday AM meeting - begin around 0845 on June 5.

• See state discussion above. [Editor's note - the discussion happened today, I added the notes where they fit logically in the Interaction section]
• Catalog 1.0 is stateful, carefully worked out.
• What are the real implications for picking WSDL?
• Proposal?
  • Use WSDL as a central part of the OWS RFQ strawman architecture

Catalog/Registry/Metadata issues

• Doug's premise - most searches are done for content and once the content is found, services are invoked. Concern about a "layer-oriented" approach once we move away from a WMS-only suite of web services. E.g. WFS, WCS, etc.
• Paul's desire - be able to add metadata at finer levels of granularity than the feature collection (dataset) level. Use GML to express it? Also FGDC, ISO, etc. content standards.
• Top-level metadata is captured in the catalog.
  • A relationship between the detailed metadata and the top level metadata should exist. At least via a controlled vocabulary. Be able to describe vocabularies in an externally accessible form.
• Names of operations should be in a controlled vocabulary. We're talking about element content, not tags.
• Action: Lou to send DEDSL to the group.
• OGC Should try to use existing vocabulary server mechanisms. Engage with GILS, UDDI communitites. Possibly, in the interim, extend the WFS to return the set of terms used in a given feature collection. - Rob (and Doug) to liase with the UDDI/GILS community.

Lou's talk - (Action: Allan to get slides - got 'em here: [ppt, pdf] )

[note - Lou put this together before we decided to get serious about WSDL, he will look at this again based on WSDL...]

Discussion on type hierarchies

Generally like what Lou has described

Don't have a clean separation of the bundling, e.g. GetMap - it's more than just portrayal right now.

Chaining - e.g. Ionic's WMS can chain to a WFS as a data source (but it's the "opaque" variety based on 19119's terms)

Even if we unbundle or refactor, people still want to give common names to these things.

Lou is looking for (a) additional resource and/or participants to take this further. Need to actually capture this into real documents...

WCS - have not heard much about this recently. There are WCS's in existence based on Gridded Coverages, GML coverage WCS are still at the "generic" level.

To really do service type hierarchies we need to move capabilities to XML Schema.

How do we discover the results of GFSPP or other pilots? One answer, plug into the existing stream of interaction. Desired answer - have some digestified results that we can read... Perhaps this can be fixed by getting the web pages better organized and making them more accessible.

Conclusions and recommendations from this meeting

1. Move to WSDL (as though we really want to use it)

1.1 Map WSDL terminology to 19119 terminology and come to grips with the differences.

1.2 Action items - try using WSDL to encode WMS, WFS, etc. using WSDL. Ron Lake volunteered to do WFS.

1.3 Could start with UML, mechanically move to WSDL. Or could dive in...

2. Move to a operation type hierarchy

2.1 What are they called and what are the relationships among them

2.2 What do we call the bundled (i.e. deployed chunks?)

3 Liase with UDDI, GILS, (Join UDDI.org? - OGC does not currently liase with UDDI.org - do it via individuals who care about it...) (action - Allan to find the report on the last UDDI meeting and publish - done [here are the notes])

3.1 Follow closely what they are producing

4 Influence on OWS RFQ and WMT-3. Need to make sure we're on firm ground.