AW: Edutella Information Integration Questions

["Bernd Kammlander" <bernd@kammlander.com>]

Some further descriptions:

- How can Edutella consumers and providers be classified
regarding design, communication and execution autonomy?

Design autonomy means that a component (one source system participating in
the federation) is independent from other systems in its design (etc. naming
concept, data model etc.). This design autonomy also entails the change of
design at any point in time, which can lead to difficulties within
infrastructure and interoperating components.

Communication autonomy is given when a component can decide independently to
which other systems it communicates. In other words it means that the system
can leave and enter the federation system at any time.

Execution autonomy means if a component is independently doing execution and
scheduling of incoming requests in a separate way. But, for example, in a
global transaction management system execution autonomy is very complicated
to retain. What is your opinion regarding the last comment?

***********************

- Can Edutella consumers and providers actually be classified regarding
technical, logical, quality-related, infrastructure, user-related metadata
and the metamodel? (relates to metadata concept using flexible metadata)

Question three deals with the metadata concept concerning heterogeneity and
distribution.
For example,
at the presentation layer (user applications) is the user-related or/and
infrastructure metadata;
at the federation layer the semantic, logical, quality-related or/and
technical metadata;
at the wrapper layer resp. foundation layer (data sources) the metamodel;

Technical metadata is used to bridge technical and interface heterogeneity
and therefore describes information regarding technical access mechanism
such as the protocol, speed of connection or query capabilities.

Logical metadata relates to schemas and their logical relationships,
especially relationships and dependencies between several schemas of one
data model.

Metamodels as metadata means the support of the interoperability of schemas
in different data models. Semantic metadata can be used for describing the
semantics of concepts. In particular, ontologies and thesauri can be helpful
to fulfil this purpose. What is your opinion in this case?

Quality--related metadata describes source-specific properties of
information systems (e. g. reliability, actuality or comprehensiveness).

Infrastructure metadata includes navigational aids like annotated bookmarks
to help users to find relevant data.

The user-related metadata describes responsibilities and preferences of
users such as user profiles.

***********************

Is it correct to say Edutella a query forwarding architecture. If not what
are the added values?

***********************

Sorry for the undefined questions, but now I hope it is clear and an answer
can be given.
Thank's in advance.

--Bernd



-----Ursprüngliche Nachricht-----
Von: Wolf Siberski [mailto:siberski@learninglab.de]
Gesendet: Dienstag, 08. April 2003 12:49
An: discuss@edutella.jxta.org
Betreff: Re: Edutella Information Integration Questions


Bernd Kammlander wrote:
> there are some questions from the informaiton integration
> background. I hope some of them can be answered.
I'll try my best

> All questions refer to a usual query process scenario, where a Swebok
> consumer and an OLR provider (simple relational database) is used.
>
> 1. How can Edutella consumers and providers be classified
> regarding design, communication and execution autonomy?
I'm not sure what you mean here.

> 2. How can Edutella consumers and providers be classified concerning
> syntactical, data model and logical heterogeneity?
Syntactical heterogeneity is not allowed. All peers communicate
using QEL (Query Exchange Language). All peers are supposed
to present their data in RDF.
The place where Heterogeneity is allowed (and expected) is
the usage of RDF Schemata. In the new super-peer architecture,
each provider sends a self-description to its super-peers which
contains the schemas used. This can be used to classify
peers by schema.

> 3. How can Edutella consumers and providers be classified regarding
> technical, logical, quality-related, infrastructure,
> user-related metadata and the metamodel?
What exactly is the difference between your questions 1-3?
It seems there is a large overlap. It would help very much
if you could give sample classifications to see what you
expect, and describe for which purpose you want to use
the various classifications.

> 4. How does the mediator in Edutella solve semantic and
> structural problems?
A mediator is not yet provided in the implementation.

> 5. Has the OLR wrapper a relational schema or explicitly
> predefined queries against its schema?
The OLR wrapper assumes that the RDF triples are stored
in a specific relational schema. There is no predefined
RDF schema or predefined queries, so any kind of RDF
data can be used with the OLR wrapper.

> 6. On which side is the wrapper built? On the data source
> side or mediator side?
If you assume the typical database mediation architecture,
then on the data source side. However, Edutella doesn't
have a mediating component yet.

> 7. Is the export schema identical to the schema of the olr provider
> database? Is it a necessity?
The export schema is always an RDF schema,
and it is the schema used for creation of the RDF model.
It doesn't have anything to do with the relational schema
which just describes how the triples are stored.
The wrapper cares only about technical translation
(between relational and RDF model rsp. query language),
but not about semantic mediation.

> 8. Is there a gateway between Edutella wrappers and data sources?
I don't understand this question.

> Thank's in advance and hopefully the questions are defined in
> an easy way.

IMHO it would be simpler to answer if you could describe what
you want to achieve. But I hope this answer contains at least
part of the information you wanted.

--Wolf