I have seen little discussion exploring architectures for an OOR.
Perhaps this
is because the requirements are still in flux. Still, I think we can
safely make
some assumptions about minimum requirements and start exploring options.
The remainder of this email does just this. (01)
*************************************************************** (02)
Open Ontology Repository Architecture possibilities (03)
Bare minimum requirements: (04)
1) Need to support ontologies in at least OWL and Common Logic (CL).
2) Need to provide means for users to Discover content via browsing
and query of exposed elements of the content and metadata related to
that content.
3) Need to serve content reliably in an appropriate standard exchange
form
and using protocols associated with each content type.
4) Need to provide persistent and available access to this content
and associated metadata for multiple versions as the content evolves. (05)
I would characterize requirements 1 and 2 as requirements on a
Registry functionality and 3 and 4 as requirements on a Repository
functionality of an OOR. (06)
Below I discuss some possible architectures for an OOR to meet these
requirements. I am somewhat familiar with ebXML RegRep and XMDR, so
below are some architectures for repositories that would use these
specifications. There may well be other, as good or better,
alternatives. If there are, people should decribe them as well, so
that we can flesh out all the alternatives and compare them. (07)
Some possible OOR architectures: (08)
A) Reuse and extend OMV infrastructure to add CL specific
enhancements. OMV already provides rich support for the above
needs (not sure about version management) for OWL based content.
This is built on an ebXML based RegistryRepository. (09)
B) Create a MOF (Meta Object Facility) profile for ebXML RegRep and
then use metamodels defined in the OMG Ontology Management Metamodel
(ODM) to store content.
Would need to:
- define models for related metadata (use and extend OMV?) which
would work across supported content metamodels,
- find or develop software to convert between MOF and exchange
forms,
- build query/browse front-end that works with content and metadata, (010)
C) Use an XMDR repository.
Not sure what is needed to extend the infrastructure for XMDR (as
described in Bruce's slides [1]) to meet the above requirements.
XMDR should be able to support a range of forms for conceptual
models including CL and OWL, but with OWL and RDF tools sprinkled
around the diagram on slide 16, I am not sure if all functions
would work on all formats. Clearly the MDR pedigree should mean
excellent support for some kinds of metadata at both the model
and model element level. But some extensions will be required to
support language specific notions like Description Logic expressivity. (011)
D) Use multiple distinct infrastructures supporting different
formats which would then be federated to appear as a single
repository via one of any of number of mechanisms (redirection,
front-ending by a single system, front-ending by a federation of
systems each of which wrap specialized systems). (012)
We probably should create a set of questions to ask about each
OOR alternative architecture, the answers to which could be used to
evaluate the alternatives. (013)
BTW - I don't think that the OWL Lite profile for ebXML RegRep should
be used for OOR. The profile provides both more and less than what we
need even for managing OWL-only content and doesn't address any FOL
content forms. (014)
BTW2 - There was some email discussion about exploiting the
relationship of OWL as isomorphic with a fragment of First Order Logic
to simplify some part of this infrastructure, say the Repository, by
supporting only a FOL form (such as CL). This could work, but I am
not sure that it buys you much. Most users and tools for OWL ontologies
will want to browse, refer to, and import OWL forms so we will have to
provide interfaces and export forms in OWL anyway. But we will have
created the task of defining the fragment and dealing with new
round-tripping
issues that the transformations between OWL and this fragment will create.
Also, will not the difference in semantics between OWL DL and OWL Full
also
complicate this approach? (015)
-Evan (016)
Evan K. Wallace
Manufacturing Systems Integration Division
NIST (017)
[1]
http://ontolog.cim3.net/file/work/OpenOntologyRepository/2008-02-28_Ontology-Repository-Landscape/XMDR-input-to-Open-Ontology-Registry_v2--BruceBargmeyer_20080228.pdf (018)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/oor-forum/
Subscribe: mailto:oor-forum-join@xxxxxxxxxxxxxxxx
Config/Unsubscribe: http://ontolog.cim3.net/mailman/listinfo/oor-forum/
Shared Files: http://ontolog.cim3.net/file/work/OOR/
Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository (019)
|