[Freegis-list] Re: metadata storage and fgdc

[Stefan F. Keller]

Dear Alfred (Alfred de Jager <alfred.de-jager at jrc.it>),

In freegis-list-request at intevation.de you wrote:

> Date: Tue, 02 Apr 2002 18:09:01 +0200
> From: Alfred de Jager <alfred.de-jager at jrc.it>
> Organization: Joint Research Centre
> To: freegis-list at intevation.de
> Subject: [Freegis-list] Re: Freegis-list digest, Vol 1 #337 - 4 msgs
> 
> Dear Alexander,
>
> Regarding your geodatabase metadata storage and fgdc metadata collectors. Please 
> take care before using these tools. They are merely conceived to collect in one 
> moment metadata not to update or relate between them. In order to do the latter 
> you better throw away all those flat-file models and start over with a
> 'normalized' model containing only the essentials. 

I completely agree!

> Hint, try to update the phone number of your 'contact person' using ArcGIS.
> The problem of the whole GI metadata hype is that we do not apply the normal IT 
> rules of data normalization (no redundancy) and essentialization (store the 
> minimum and derive what is derivable). 

That's what we also try to do: take the most important thing of geodata,
that's the geodata schema (precisely described in INTERLIS (=>UML=>XML)
and derive from that what you can for metadata purposes. There is also a
object catalog (data dictionary) needed (in XML; there are papers on
this methodology, see e.g. http://www.integis.ch, some are also in
english available). The object catalog and the remainder needs
definitely to be defined in a metadata/catalogueing model. That what's
ISO 19115 is doing. 

As a matter of fact, there is no sw around (I know of) to do this
derivation and metadata management job.

> GI Metadata is still an enduser wish list 
> and in any implementation I have seen thus far the whole thing is mixed and end 
> user perspective driven. Apperently nowbody cares about the managebility of 
> these data! I think we now know what the end user wants! We have around 3 
> official committees writing down cryptic definitions for field names like 
> 'title', 'author' etc. What we now need to do is to define which information is 
> derivable and which is not. From what is not derivable you need to check if it 
> is essential (is my faxnumber essential to know at the level of metadata?) if 
> not skip it.
> 
> The remaining information needs to be normalized and stored in a database out of 
> which you generate (perhaps xml) information sheets. One does not store XML 
> sheets!!!!!!! 

Why not? XML is always better than .doc-Files, i.e. for explanatory
text! An also better than ORACLE dumps. Depends what is in the XML!  

>Normalization means that every repeated piece of information (eg a 
> contact person) results in a separated relatable table. Such a database does not 
> exist in a generic way since the essentialization must result for any 
> organization in a slightly different model. To exchange information between 
> organizations both organizations must not only agree on a physical protocol like 
> Z39 but more on the identification of their objects (contact persons, 
> organizations, placenames, keywords, reference systems and the whole bunch).
> 

I again completely agree! This is exactly why we developed INTERLIS for
defining geodata and data models. In addition - as a free side effect -
we can derive formats automatically - even GML if this is needed.
Currently GML is not enough so that we use an own INTERLIS/XML-format
(see www.interlis.ch for free documentation, ask interlis at lt.admin.ch
for an official englisch version of this standard). 

> Conclusion.
> Metadata systems when seriously applied do not come out of the box, you ought to 
> develop yourself a model and if you what to exchange metadata with organizations 
> outside your model you should first agree on object identifiers before even 
> bothering on exchanging.

With this INTERLIS geolanguage we are able to define and exchange even
metadata and catalogues. There is a working group to define a Swiss
profile if ISO 19115 Metadata! There are some papers on this at
http://www.kogis.ch.

You have mentioned FGDC and ArcGIS/ArcCatalog. These models are
outdated! We recommend the ISO Core (or even a subset of if like you
suggest). Even OGC and ESRI is expected to change to ISO Metadata
(because the project leader of the ISO group moved to ESRI, as I heard
from my ISO expert colleagues!).

> My Hints:
> Most databases contain a 'comment' field for any object, unstructured metadata 
> (definition of what it is, what units are used) can be stored best in there, 
> close to the data. For organisational information you just use the contact 
> administration of e.g. your browser in which an email address can be an unique 
> identifier. Most browsers allow in various modes to 'share' that information and 
> make it searchable. Note that browser contact information is not normalized. 
> Thus if your organisation is large you better buy a contact administration 
> package. Most GI metadata is derivable from the dataset themselves and using 
> simple batch programs you should be able to make that searchable as well. 
> Geodatabase does the latter, partly, and if convenient for you, you can use 
> that. In Oracle (and other databases) the whole database structure is stored in 
> the database dictionary and thus using Geodatabase on top of Oracle is a little 
> bit double (understatement).
> 
> What then remains is the creation of one table containing.... two fields:
> 
> "contactperson email", "physical dataset location"

Perhaps a little too small list (:->)? I'd suggested in addition a
"prize indicator". 

> Just imagine the simplicity.

The message of N. Wirth and thus of INTERLIS was similar: "The Art of
Simplicity".

> Ciao
> --
> Drs. Alfred de Jager
> Geographer, GIS and Database Consultant
> 
> Portal for European Landscape Research
> http://www.aris.sai.jrc.it/en/search-tools

Regards
-- Stefan Keller
___________________________________________________________________
Prof. Stefan F. Keller, Dozent fuer Informatik
Center für integrierte Geo-Informationssysteme (int>e>gis)
am Institut ITA-HSR der Hochschule fuer Technik Rapperswil (HSR)
CH-8640 Rapperswil, Tel. 055 222 47 46, Fax 055 222 44 00
mailto:sfkeller at hsr.ch, http://www.hsr.ch und http://integis.ch
___________________________________________________________________
HSR, der schönste Campus der Schweiz... siehe http://netcam.hsr.ch/