RE: [tm-pubsubj] New Gentle Introduction

["Bandholtz, Thomas" <TBandholtz@koeln.sema.slb.com>]

Title: RE: [tm-pubsubj] New Gentle Introduction

something about human vs. machine:

I like Steve's images, but something is missing:

1. The human actor is always the actor, even if she uses a machine to interprete something machine-readable. The human actor is not named "human interpretable", as it looks in the pictures :-)

2. For the notation of PSI Steve uses http://psi.fruit.org/#apple (fragment identifier "#") and http://www.everything2.com/index.pl?node_id=187149 (query "?"). Both are valid addresses (exactly: URI references), so you can use them as a PSI, but there is a difference. We had this discussion last year. For those who are not familiar with this difference, please ask, or simply forget it.

To make it more clear, I compare http://psi.fruit.org/#apple with http://psi.fruit.org/?psi=apple. Poor fruit.org would not be able to handle this. Either you have a formated document in this place containing an anchor (or XML-ID) with name "apple", or you have a process (such as a Java servlet) interpreting the query (psi=4711). Why should they be able to support both?

This is not because of the syntactic problem of "#" vs. "?". In the standards (XPointer) there is no difference, as http://psi.fruit.org/#apple is specified to be a short cut for the XPointer notation of references to the XML node with xsd:ID="apple". But this is not commonly implemented. Nothing has changed here since last year's spring.

But, even if XPointer would be implemented, this would not solve the problem. The human reader wants to see a formated document - the machine needs pure XTM.

No problem to use the same XTM source for both of them, and have it XSLT processed for the human reader. But how can poor little fruits.org decide how to present the apple if there is only one URL for both? No chance :-(

It is only the human who can decide what he wants. If fruits.org support machine readable XTM only, the human has to use a machine (f.e. XSLT interpreter) to get the formated document. But all the formating then depends on the specific machine and XSLT script. Both may damage the semantics of apple.

Fruits.org wants to control the formating process. But it also wants an "intelligent agent" in the semantic web to access the pure XTM. I really do not know how to solve this without two (2) URL for each PSI: http://psi.fruit.org/human.html#apple and http://psi.fruit.org/machine?psi=apple. Both say that there is a subject with ID apple published by fruits.org, but there are different presentations.

This could be handled by something like an extension of XML namespaces that could look like:

<psi:exmlns psi:prefix="fruits" psi:html="http://psi.fruit.org/human.html" psi:xtm="http://psi.fruit.org/machine"/>

Then you could use "fruits:apple" in the following document freely, and the interpreter can decide which URL to use. Isn't this an issue for XRI, Bernard?

But this is dreaming about the future. We have to find a working solution now.

The first step is to completely separate the semantics of PSI from the syntax.

The second step is to clarify the process: It is not the human who reads the HTML code, but he reads the formated document. The machine is always in the game - be it a standard browser processing the HTML code into the readable formated document. It is a fact that browsers are starting to learn XML, so I believe that we will be able to support only the XTM presentation in the future.

Thomas