Hi John,
Thanks for this very useful feedback! I have not heard the concern during the OGC meeting where I was, but metadata was discussed during many sessions. I can't remember Simon mentioned it during the ebRIM discussion and I can't remember anyone else mentioning it... In fact the lack of a free text search option could be a mayor constraint. I there no way around this issue? I can also see benefits in the ebRIM approach and have thus not made objections to the proposal. The free text issue is definitely one that needs further discussion before a final decision is made.
We have not decided anything on the ebRIM profile yet as it all needs to be discussed further at OGC in anyway. For now we will concentrate on the ISO profile and most likely also on the ESA proposed Earth Observation profile (implementing client support to EO profile based catalogs). Any ebRIM work would I think for sure not happen before next year and from your message I can see that there's still enough to be discussed before really moving towards ebRIM support.
Thanks a lot,
Jeroen
On Jul 18, 2006, at 8:39 AM, John.Hockaday@anonymised.com wrote:
Hi Jeroen,
I hope you don't mind me taking this discussion directly to you. If you like
please send it to the GeoNetwork developers lists.I have noticed that you are on the OGC ISO 19115/19119 revision working group
and that you have a great deal to do with the development of GeoNetwork. As
you know I am very involved with the implementation of ISO 19100 standards.
I expect that you are aware of the Edinburgh OGC recommendation that all
Catalogue Services - Web (CSW) should be implemented using ebRIM. I notice
that GeoNetwork doesn't use the ebRIM implementation.GeoNetwork stores the metadata as a single object in a RDBMS. This is great
because it means that the essential free text search can be done on the
column of these objects. This search is covered by the "anyText" search
element of CSW. I have found from my 12 years of management of spatial
search metadata systems that the majority (~80-90%) of searches are free text
on any component of the metadata. That is, the equivalent of "anyText" is
used for the majority of metadata searches.However, ebRIM takes a subset of the metadata and stores this in RDBMS
tables. Searches are then done directly on this subset. Once the metadata
is identified by the search results of this subset, the metadata record can
be presented to the user.ebRIM does not implement the "anyText" and this is a major flaw of the ebRIM
CSW implementation. It is likely that a majority of the metadata content is
not accessed for search results and therefore it will not return the correct
hits expected by the users.A data custodian is not likely to fill out an optional metadata element
unless they feel that the element is essential to describe their data. If
they identify this importance then it is also important that this element be
available for users to search. If ebRIM does not index these optional
elements than it will not be meeting the expectations of the data custodians.I have voiced my concerns to Simon Cox who I believed tabled them at
Edinburgh. However, this was voted down and so any implementation of ebRIM
for CSW is likely to not meet the users existing expectations and needs.If GeoNetwork was to adopt the ebRIM implementation of CSW then I believe
that a major part of GeoNetwork's functionality will be lost. Do you expect
that GeoNetwork will adopt the ebRIM implementation of CDW? If so when will
this work be done?How do you feel about this reduction in essential usability of GeoNetwork?
Thanks.
John Hockaday
Geoscience Australia
GPO Box 378
Canberra ACT 2601
(02) 6249 9735
http://www.ga.gov.au/
john.hockaday\@ga.gov.au