RE: [xdi] Rationale and semantics for the 3rd markup model

Ajay and Loren,

 

First, Ajay, I agreed with all your points below. I think your last
sentence, "I would think [these terms] could be used to describe the
independent (from the underlying data sources) model" captures the key
distinction we are trying to decide on the best semantic term for, i.e., that
the third model is independent from any specific schema that may be used either:
a) by the underlying data sources (i.e., native database or directory schemas),
or b) in conventional XML markup.

 

This jibes with what Loren
suggests in his message: "How about the "Schema-less" or
"Semantic" or "Self Describing" models?"

 

I think
"schema-independent" is better than "schema-less", because
there is still a schema involved (the native XDI schema). And although I
understand Loren's suggestion of "Semantic", I think we should avoid this
term to prevent confusion with the Semantic Web. 

 

"Self-describing"
is also a fascinating option. My first reaction was to say that any type of XDI
markup is self-describing (indeed, I see it often said that XML markup is
"self-describing"), so I didn't think its as good as
"schema-independent" in capturing the essence of what distinguishes
the third model. However the more I thought about it, the more
"self-describing" might provide the right contrast to
"inline" and "enveloped", because
"self-describing" suggests that understanding the data does not
depend on understanding another XML schema, and indeed that is the key
distinguishing feature of the third model.

 

The only other suggestion
I would make, along the lines of keeping it short and intuitive, is that a key
benefit of schema-independence and self-description is "generalization".
In other words, what the third model gives you is the ability to do generalized
data sharing across any two applications or domains regardless of underlying
schema. Following this logic, the third model could be called the "generalized
model" or even just the "general model'.

 

So, we're narrowing it down
to three candidates:

 

* Schema-independent

* Self-describing

* General

 

Preferences? Other thoughts?

 

=Drummond 

 

 

From: Ajay Madhok
[mailto:] 

Sent: Tuesday, November 30, 2004
3:27 AM

To: 'Drummond Reed'; 'Loren West';
'XDI TC'

Subject: RE: [xdi] Rationale and
semantics for the 3rd markup model

 

Hi Loren and Drummond,

 

My two cents on the ‘Logical’ Model as I
saw my name associated with it:

 

XDI is a paradigm shift that will allows us to replace
EAI (Physical
interconnection of systems to enable data aggregation and exchange) and other
middleware (for Physical shuffling
and translation of data between different systems) without having any semantic,
procedural or structural knowledge of the underlying data sources. The end goal
is a logical, abstract, distributed Data Service which does not have to deal
with the context of each underlying data source.

 

This said, I do agree with you that the words I have
used here – logical, abstract, distributed are ‘universally’
over-hyped and. I would think they could be used (at some risk of coming across
as marketing fluff) to describe the independent (from the underlying data
sources) model.

 

More on the call.

 

cheers

=Ajay

 

 

 

 

-----Original Message-----

From: Drummond Reed [mailto:] 

Sent: Monday, November
 29, 2004 11:19 PM

To: 'Loren West'; 'XDI TC'

Subject: RE: [xdi] Rationale and semantics for the 3rd markup model

 

I see your point, and I agree it means
"logical" is probably not the best

name for the third XDI markup model, because
"logical" could equally applies

to the abstract XDI model as a whole, not to any one
markup model.

 

In that case I believe the single most distinguishing
characteristic of the

third model is that it is independent of any XML
schema OTHER than the XDI

schema. This is the biggest contrast with the other
two models, which are

both XML-schema-centric. In which case the most
precise label for the third

model would appear to be either "the independent
model" or "the XDI schema

model".

 

=Drummond 

 

 

-----Original Message-----

From: Loren West [mailto:] 

Sent: Monday, November 29, 2004 12:44 PM

To: 'Drummond Reed'; 'XDI TC'

Subject: RE: [xdi] Rationale and semantics for the 3rd
markup model

 

I'm not a big fan of the terms logical, physical, and
abstract

because they don't communicate anything in and of
themselves.

Sort of like the word "low" - it only makes
sense when compared

to something "high".

 

If one model is logical, what is it compared to? 
Illogical?

Physical?  And what is it that makes it more
logical than the

others?

 

ALL 3 models allow you to identify, describe, and
exchange data

so none of them are any more or less XDI than the
other.

 

I think it would be a mistake to say that the inline
and enveloped

models are just a translation of a logical XDI model
into XML

because it places emphasis on the model itself vs. the
virtues

of identifying, describing, and exchanging data.

 

It suggests that you should understand the model
first, then 

understand it's translations into XML if that's how
you want

to use XDI.  I fear we'll lose people if we try
to get them 

to understand the 3rd model in order to understand
XDI.

 

=Loren

 

-----Original Message-----

From: Drummond Reed
[mailto:] 

Sent: Monday, November 29, 2004 11:37 AM

To: 'XDI TC'

Subject: RE: [xdi] Rationale and semantics for the 3rd
markup model

 

 

Loren,

 

I think I get what you're saying - that the two
applications that need to

exchange data may need to first establish context, and
that's what this

"meta-context" is for.

 

But in that case, as you said on the last call, the
application *is* XDI.

XDI is the context.

 

That's why it strikes me that most natural name for
the third model is the

term Ajay uses - "the logical model" -
because the goal of XDI is to provide

an abstract, logical model for identifying,
describing, and exchanging data.

All forms of XDI markup, including inline and
enveloped, serve to map data

to and from this logical model. But the first two do
so from the context of

existing XML schemas. The third one does so from the
context of the logical

model itself.

 

=Drummond 

 

-----Original Message-----

From: Loren West [mailto:] 

Sent: Monday, November 29, 2004 11:12 AM

To: 'Drummond Reed'; 'XDI TC'

Subject: RE: [xdi] Rationale and semantics for the 3rd
markup model

 

Hi Drummond,

 

This is the best description of the 3rd model
yet.  It's still

a stretch of my imagination to think about an
implementation of

software that could take advantage of this format,
because

by the time I'm thinking about implementing software,
I'm in

context, and I don't know how to handle out-of-context
data.

 

I can imagine, however, a conversational application
where

context is communicated in real-time - much like human
conversation -

and data is retrieved and delivered based on ever
changing context.

 

This would be a great way to communicate in that
application.

With this said, I'd like to suggest a few more names
for it:

 

Runtime model

Real-time model

Conversational model (I like this the best)

 

=Loren

 

 

 

-----Original Message-----

From: Drummond Reed
[mailto:] 

Sent: Saturday, November 27, 2004 2:03 PM

To: 

Subject: [xdi] Rationale and semantics for the 3rd
markup model

 

 

I had an action item from Wednesday's XDI TC call (for
which Marc is still

working on the minutes) as follows:

 

# Start a thread on the list to: a) solicit our
collective rationale for

using each of the three XDI markup models, and b) out
of this, develop

consensus on the name we should use for the third
model. 

 

The three XDI markup models were summarized in

http://lists.oasis-open.org/archives/xdi/200411/msg00041.html.
The first two

models are both "XML-centric", i.e., in both
cases XDI markup is added to

existing XML markup in order to gain the data sharing
virtues of XDI while

still leveraging any existing investment in XML
schema, tools, and

applications.

 

The only difference between first two models comes
down to whether the

existing XML schemas being marked up with XDI are
extensible or not:

 

1) IF THE EXISTING XML SCHEMAS ARE EXTENSIBLE, then
XDI markup can be placed

directly inline with the existing XML. Thus the
consensus on the call was to

call this XDI markup model "Inline".

 

2) IF THE EXISTING XML SCHEMAS ARE NOT EXTENSIBLE,
then XDI markup must be

added in an envelope wrapping the existing XML. The
consensus on the call

was to call this XDI markup model
"Enveloped". 

 

Because these two models are so clear, most of the
discussion on Wednesday's

call focused on the third model, and when and why an
implementer would want

to use it. Three rationales were put forth:

 

  1) When richer resource description is needed
than is available using XML

QNames (and thus is better provided by XRIs).

 

  2) When the data being marked up does not yet
have an XML schema.

 

  3) When the data being aggregated in a single
XDI document does not lend

itself to a particular XML schema because comes from
many different data

sources and there is no single XML schema that makes
sense for aggregating

all of it.

 

After the call, I realized the killer example of the
third type of data is

when the resource being described is one as
generalized as "person",

"organization", or "topic". It
would be all but impossible to create a

single XML schema that can describe all the data that
might be associated

with a person, an organization, or a topic. The reason
is that these are

simply very general concepts, that can be reused in
thousands or millions of

data sharing applications.

 

It dawned on me that this made perfect sense. The
universe of XML schemas

today exist mostly to describe data in particular
contexts, i.e., existing

applications and data stores that already have their
own schema. Moving the

data in XML in this case is the first step in being
able to share it with

other applications, domains, and Web services.

 

But the more generalized the data, the harder it
becomes to completely

satisfy this need with conventional XML, because the
data is less and less

"schema-specific" and conventional XML is
designed to put data in a specific

schema context.

 

It is for these classes of applications, where data
may be reused across

many (i.e., hundreds or thousands) of different XML
schemas, that the real

demand for XDI arises. Of course all 3 markup models
supply the solution:

XDI markup can identify and describe the data in a
schema-independent

manner, so the same logical data can be identified and
shared across many

different physical XML schema instances. But if the
distinguishing criteria

of the first two models is that the data already
exists and makes sense to

be used in the context of existing XML schemas, then
the distinguishing

criteria of the third model is that either:

 

a) No XML schema currently exists to provide the data
context, or 

b) The context is so general that no single XML schema
CAN exist to provide

sufficient context.

 

Although in the first case it can be argued that the
easiest route to

establishing the context is to simply create a new
conventional XML schema,

in the second case that's not an option. There is no
such schema. The only

solution is a "generalized" or
"universal" XML schema that exists to

describe things independent of any one specific XML
schema context. That's

the third model.

 

Loren predicted that the right term for the third
model would emerge once

the rationale for the third model was clear. I would
summarize the

rationaled for using the third model as "the
schema to use when you need

express and reuse data in a generalized, universal, XML-schema-independent

format". This would suggest the following names:

 

General model

Universal model

Independent model

Schema-independent model

Metaschema model

Abstract model (Victor's suggestion on the call)

 

Votes? Other suggestions?

 

=Drummond