[Freegis-list] Re: Metadata revisited - answers to Alfred de Jager

[Stefan F. Keller]

(Thanks too to all of you for this discussion!)

Mr. de Jager

Many important issues you raise in your last mail 
(Thu, 04 Apr 2002 13:30:15 +0200):

> the links you send for implementing a Swiss GI infrastructure
> my main wonder was if this is aiming at the swiss context only or 
> are you interested to go beyond that?

First (10 years ago), there was the observation, that geodata is at the heart of
any infrastructure and that the main problem of any (!) interoperability
(including data transfer, online and metadata services), is communication. This
is done through a precise, both human and computer readable data schema. Thus,
INTERLIS ("the GeoLanguage") is mainly a communication language. Then, INTERLIS
was implemented in SW since five years. Now INTERLIS-based SW is first in
showing ISO-concepts.

INTERLIS has also been rated first in an internal ISO evaluation but was
penaltied because of lacking tool support (that was years ago), lacking
graphical notation (that now is UML), perhaps some lacking politics/lobbying.
Now, recently Switzerland (and INTERLIS) got a second chance becoming an ISO WG
leader for a working geometry profile. Interesting enough, OMG (www.omg.org) -
the issuer of UML and CORBA - now is changing its fundamental approach to
interoperability to "model based architecture", which is the same in principle
as I try to explain here.

> The INTERLIS language seemed to me more a translator. Useful but (...)

It is a unfortunately broad misunderstanding what INTERLIS is: It's not a SW -
it's an official Swiss standard body paper specification (freely accessible in
de/fr/it/en). It's not a cadastral format - it's mainly a schema description
language... and additionally it is combined with transfer format rules, which in
turn define an exchange format (this mechanism can be extended to interface
specification generation (for online APIs), like IDL).

Important last remark: One has to keep in mind that modeling is a demanding
tasks and (even GIS project leaders in governement who are data owners) IT
skills are often lacking - as you too seem to state.

> (...) more a translator. Useful but I pinpointed that the main 
> issue of data transfer is unique identification of normalized 
> objects (...)

I just can't follow, why you would like to organize IDs for (always) every 
object for metadata: first what is this ID-linking for? If you organize mainly a
portal/clearinghouse and if you agree, that metadata on its own is not a large
dataset, then wy not simply overwrite existing datasets? Second, you probably
know that there are not many GIS around yet (I know of one or two in
Switzerland) which can manage IDs: you have to consider their deep implications
in the core db.

Having said this, I's like to repeat, that we got a proposition for generating
world wide "Simple OIDs" (SOID or simply OID) coordinated with german cadastral
authorities (see www.interlis.ch > Services where also the INTERLIS 2 spec are
downloadable).

The difference I pointed out between "normalized" objects and "object-oriented
robust modelled" objects we can discuss perhaps in another thread.

> Thus if we want to define any organization as a data source than we 
> ought to make a world wide web service giving a unique id to any 
> organization that can be a data source (...)

That is also feasible with the SOID: there, the idea was, that every data source
owner can decide what is exactly identified: whether it is a GIS, a single
database-server component/module or every SOID generation component/module.

> Storing XML: (...)
> (...) Just store the fields (...) as normal strings, dates and number
> in your database and make searching and update fast by default.

There remain two problems one of defining APIs for (cascaded) online-queries and
the another problem of exchanging/archiving metadata. The latter we think having
solved (using the INTERLIS-mechanism which is usefil for any schema). For the
first problem (assuming that this not solved yet a OpenGIS...) it seems to me
most interesing to do more research, discussion and stndardization in looking
for Napster-alike peer-to-peer concepts leading hopefully to uesful and tested
protocol specifications.

> ...

Regarding your last suggestion for a "free" OID service, I would like to forward
this to some of my Swiss colleagues - perhaps we could offer such a thing in a
short time. 

The main question remains to agree on the goal of such OIDs and to agree on a
OID specification (like the SOID proposal).

Regards 
-- Stefan Keller
___________________________________________________________________
Stefan F. Keller, Professor für Informatik
Center für integrierte Geo-Informationssysteme (int>e>gis)
am Institut für Internet-Technologien und 
-Anwendungen (ITA-HSR) der Hochschule Rapperswil (HSR)
Oberseestr. 10, CH-8640 Rapperswil
mailto:sfkeller at hsr.ch, http://www.integis.ch und http://www.hsr.ch
___________________________________________________________________