[Sigia-l] CMS and IA

[Nuno Lopes]

(Warning long post)

Hi there, I'm very interested in this subject. I and a couple of friends
we are "trying" to build a CMS encompassing our experience in the past
12 years. My self, I've been developing custom CMS solutions and
products for many different set of industries spanning from
Pharmaceutical Industry and Telecommunications to Interactive TV
broadcasters.

Our discoveries and research, and actual implementation will be released
to a community with common interests and expertise for free.

We are in the modeling phase and making some prototypes to test the
feasibility of different approaches encompassing the state of the art of
technology. We are building it using the .NET Framework and a RDBMS to
store and retrieve data for the moment. This was a decision due to more
of practical reasons of our own than anything else.

The novelty of the system (if there is something new about it) is that
users with minimal technical expertise (help of user interfaces and
wizards) will be given total control of its behavior and not to mention
information models. Also from the point of view of architecture we will
try to intertwine the best practices in Information Modeling,
Accessibility, Management and Security seamlessly within a systemic
environment.

I've been reading articles about Faceted Classification, Ontologies,
Taxonomies, etc etc. Technically speaking and from the engineering point
of view I come from OOA and Relational Modeling. I have to confess that
my "ignorance" of the topics that are often talked on this list lead me
to observe this articles with certain skepticism around their novelty.
In other words, most of the topics being addressed I see equivalences
with Relational Modeling and OOA at some point (even between them). With
leads me to suspect that probably they don't solve much more then what
we already have mainstream (RDBMS, OODMS, etc) from the point of view of
engineering.

Back to our system, we have divided the different features of the system
in 6 subsystems in the following manner:

modelingSys - The Modeling System is responsible for handling the
creation, management and content structure compliance by defining
information schemas that set the structure and "static" meta-data of
human readable information (structured and unstructured information that
a human can read).

workflowSys - The Workflow System is responsible for handling the
creation, management and execution of work flows over content modeled
with the modeling system.

publishingSys - The Publishing System is responsible for handling the
organization and deployment of content in multiple formats to multiple
devices. Within this scope it offers several types of publications
structures (organization) targeted to help publishing full web sites,
from a single page in a book or corporate brochure to a full blown paper
based publication.

findabilitySys - The "Findability" System is responsible for handling
the creation of multiple strategies aimed to facilitate the search of
information according using well known multiple organization patterns
(Indexing, Categorization, Faceting, and others).

storageSys - The Storage System is responsible for the creation and
management of physical storage space on multiple storage devices,
including Relational Database Management Systems (for now only in the
RDBMS in the File System).

securitySys - The Security System is responsible for handling the
creation and  management of user access profiles set against content
management operations according to their roles. Furthermore it handles
the authentication of users according to security constraints has set in
the defined security profiles.

Apart from our experience, to guide the design of features we are using
the book Content Management Bible from Boiko.

So my first question is - what are the main differences between
Relational Models and Ontologies apart from the existence of the notion
of Classes and Class hierarchies?

This leads me to the next question, what is the main difference between
"Ontology" and an Object Oriented Data Model defined in UML for
instance? 

Are the so called Slots not more then Class attributes? Is it that in a
classical OODM we can extend or restrict Class attributes and Classes in
the scope of Ontologism we may only restrict the possible values of
Slots in the sub classes?

To understand what I'm saying take the classic "Wines" example. In this
sample there is always a root called Wine. It seams to me that this
class defines all the properties (Slots?) that are available to its sub
classes. The sub classes only restrict the possible values of the Slots.


For instance, the Class Wine defines Color and Region as Slots (object
attributes in OODM). It seams to me that sub classes only fix values for
this attributes. Below the Wine Class one has Red Wine Class (just Fix
Red on Color Slot), White Wine Class (The same thing), White Burgundy
Class (The same thing but also fixes Region to Burgundy). Within this
scope aren't these Sub Classes nothing more then information views over
the extent of a more generic OODM that merely define common abstract
properties without fixing values for attributes for each class (Class
Wine only)? 

This notion of views is somewhat also linked with the notion of Views in
Relational Database Modeling. This leads me to the next question. Is
there any link between Views in Relation Data Modeling and Faceted
Classification?

I understand that all these can be viewed as different representations
of Knowledge but what I'm looking for is equivalences and differences
between each of them. In other words I would like to diminish the
influence of the linguistic terms of the IA culture or my own and tap in
to the pure truth of knowledge and comprehension behind these concepts.

In our system the user can create Content Templates (in the
modelingSys). A Content Template just sets the structure of similar
Contents but not how it looks. This is called a Template because it just
a possible interpretation of a semantic group within the model of
discourse on the system, so it may not map into a Semantic Class (also
users without an IA backround easily relate this with forms). One can
look at Content Template as a Root Class in ontology can't we some
times? Also the user can draw associations between Content Templates
(Slots Instances?). The Template for Wine information can be linked with
the Template for Region Information. Also users over this Content
Templates can create Views that basically restrict the domain to wines
sharing common properties (Color="Red"). These views can be seen as
Categories or Facets (more on that following)?

I would like also more information about how to implement a Faceted
Classification system using RDBMS. We already have the notion of Views
that basically can cut a set over a specified facet (Color, Region or
anything else). With this we could of course provide user with a way to
organize these Views hierarchically (Something managed by the
findabilitySys):

Wine:
 Color: Red, White ...
 Region: ....
 Etc...

This information could then be exported in some XML dialect to be used
by indexers or any other system.

Is this Ok? 

I'm really a noob on IA concepts and jargon, so pardon me my ignorance.
Also can you provide me with more links regarding the themes being
debated?

One other thing that I'm very interested is topic maps and how they
related with Faceted Classification schemes if there is any relationship
between them?

I know this is a long post, so if you reached here I thank you for your
patience and time.

Thanks in advance for any help.

Best regards,

Nuno Lopes