Re: meta information
Liam Relihan <relihanl@ul.ie>
Errors-To: listmaster@www0.cern.ch
Date: Fri, 3 Jun 1994 12:00:44 +0200
Errors-To: listmaster@www0.cern.ch
Message-id: <Pine.3.05.9406031026.B28562-c100000@itdsrv1.ul.ie>
Errors-To: listmaster@www0.cern.ch
Reply-To: relihanl@ul.ie
Originator: www-talk@info.cern.ch
Sender: www-talk@www0.cern.ch
Precedence: bulk
From: Liam Relihan <relihanl@ul.ie>
To: Multiple recipients of list <www-talk@www0.cern.ch>
Subject: Re: meta information
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
Content-Type: TEXT/PLAIN; charset=US-ASCII
Content-Type: TEXT/PLAIN; charset=US-ASCII
Mime-Version: 1.0
Mime-Version: 1.0
On Fri, 3 Jun 1994, Nick Arnett/Multimedia Computing Corp. wrote:
> Perhaps we aren't all looking at the value of the Web in the same way. I
> don't think that semantic tagging is an application-specific capability; it
> appears to be to be utterly essential to the Web's scalability. If it's
> not enabled, it's going to be very hard to find things.
>
> Without the possibility of semantic tagging, there's no way to automate the
> linking of Web documents in ways that take advantage of its capabilities.
> We'll be stuck with the old flat-file and relational models (which have
> already reared their ugly heads in this discussion) that have limited value
> in this environment. They have their place, but unless they're storing
> semantically related information, they'll have a hard time coping with the
> scale of information-sharing that's becoming possible.
>
> I'm bothered by the introduction of the relational model into this
> discussion. It's much too limiting. The relational model doesn't work for
> a few reasons. In a relational model, the number of sets of objects is
> basically fixed when the database is created. You can't easily add
> infinite new "columns." As I suggested earlier, in a semantic structure,
> you can add "fizzbins" any time.
>
> The second big problem is tht in the relational model, there's no way to
> express an object that inherits characteristics of two parents. For
> example, if you have manufacturing companies and service companies as one
> way of categorizing, and public companies v. private companies as another,
> the relational model has no way to express the full set of relationships.
> You have to express it as manufaturing companies that are public or private
> and service companies that are public or private, or the other way around,
> but not all of the relationships.
Agreed... complex documents (documents built in a recrusive tree
structure) such as HTML documents cannot be intuitively represented in
RDBs...we need semantic-type data models.
Unfortunately databases constructed according to these kinds of data models
tend to be filled with cross references and links....in fact so many of
them that they *could* be very difficult to maintain manually.
Therefore, if we are interested in data integrity, we need to consider
building some kind of a database level (like HAM) into the web. Such a
database level would accept and serve document objects and maintain the
integrity of the links between them.
While I suspect that tables would be easier to manage manually, they are
not rich enough for the kind of document we are interested in.
Lets face it, data integrity is a big problem at the moment (links to
non-existent documents, etc), and it is a problem that can only worsen.
With all this talk of collaboration, complex semantic relationships etc. it
look like we are moving towards consideration of a layered hypertext
architecture. Development of such an architecture would probably prove to
be one of the most challenging internet development programs ever.
However, I reckon it is worth considering :-)
Just musing :-)
Liam
--
Liam Relihan, | |\ Voice: +353-61-333644 ext.5015
CSIS, Schumann Building, - | |_/ - Fax:+353-61-330876
University Of Limerick, |__ | \ E-mail: relihanl@ul.ie
Ireland. http://itdsrv1.ul.ie/PERSONNEL/lrelihan.html