RE: [xml-dev] What makes a data component good forstandardizing?

["Cox, Bruce" <Bruce.Cox@USPTO.GOV>]

Thanks, Roger, for summarizing my five paragraphs in your one sentence.
:O) 

Bruce B Cox
Manager, Standards Development Division
OCIO/SDMG
571-272-9004


-----Original Message-----
From: Costello, Roger L. [mailto:costello@mitre.org] 
Sent: Thursday, January 15, 2009 6:27 AM
To: xml-dev@lists.xml.org
Cc: Costello, Roger L.
Subject: RE: [xml-dev] What makes a data component good for
standardizing?

Awesome message Bruce! Thank you.

There is so much information packed into your short message. I'd like to
focus today on one part. I'd like to verify that I have a correct
understanding of the "unanticipated user."

What is the "unanticipated user?"

EXAMPLE

Suppose you deploy a web service to provide currency exchange rates - a
user sends two currencies (e.g., USD and CAN) and your web service
responds with their exchange rates (1 USD = 1.23 CAN, 1 CAN = 0.90 USD).

The "unanticipated user" means the users of your web service may be
unknown to you. It does not mean that requests made by users are unknown
or arbitrary. For example, your web service can't be expected to handle
requests from a user wanting metric-to-English conversion rates. Rather,
your service only handles requests for currency exchange rates, and the
users issuing the requests may be unknown.

"Design your system to support the unanticipated user."

I think this is misleading. It should be:

"Design your system to support the anticipated but as yet unknown user."

Do I have a correct understanding of the meaning of the "unanticipated
user?"

If so, I now understand why my example of the nuclear physicist
unexpectedly sending me a document on subatomic physics and expecting me
to process it, or pieces of it, is not a reasonable expectation. Right?

/Roger
________________________________________
From: Cox, Bruce [Bruce.Cox@USPTO.GOV]
Sent: Tuesday, January 13, 2009 8:40 PM
To: Len; Costello, Roger L.; xml-dev@lists.xml.org
Subject: RE: [xml-dev] What makes a data component good for
standardizing?

Roger, your example is weak.  If the name is not useful outside its
context, then neither is the address.  Is it a shipping address? Home
address? Birth place? Current residence? Future residence? For sale? For
rent? Target? Party house?  Yes, you can put it in your contact list,
but you can do that with the name as well, and with just about as much
value.

If I understand him, I agree with Len, what is or isn't worth
standardizing is negotiated with at least a surrogate for the
anticipated but as yet unknown user.  Who buys after-market motorcycle
parts?  You might not yet know the person, but you know the role they
play in the transaction you build into your system based on the
established market dynamics that you've studied carefully and documented
fully for your SBA loan approval.  That's the context you use to decide
what to standardize.  Without that, you don't even have a market, and
therefore no competitors or related services, so why bother
standardizing anything since there aren't any other business services to
interoperate with.

Some data clumps have a context larger than a specific market segment,
such as name and address and credit card info.  They represent the
social infrastructure that makes interoperability in a specific market
segment possible.  A particular user (role player) might be
unanticipated, but the role they play is not.

The value in standardization is in reducing friction in a transaction,
which in a market is measured by changes in profit.  In an academic
context, standardization might mean reducing effort in getting
published, or the cost of publication, all tending to improve one's
standing in the community.  In every case I can think of, the decision
to standardize or not would be a cost vs. benefit for effort involved.
If the data already has a standard representation, someone has answered
that question in the affirmative.  They've done so on the basis of some
abstract model of the transaction, probably derived from historical
records, regulations, custom, or other past experience.  Surely, some
such decisions are just wrong, due to poor analysis or based on poor,
tacit, or seat-of-the-pants models.

What to standardize for the sake of interoperability is based on the
model of the operation, which has to include all the roles for potential
users, even in the hypothetical nirvana of the semantic web.  That stuff
just pushes it to a higher level of abstraction, making it all the more
difficult to achieve since there will be less chance of agreement on the
model and (I'm going to really stick my neck out here) absolutely no
chance of basing the model on any actual human behavior.  How can we
expect people to behave in accord with an abstract model that is
completely foreign to them?  Does this mean that the semantic web
actually makes interoperability more difficult rather than less?  Hmmm.
But that's completely off topic.  Sorry.

Enough rambling.

Bruce B Cox
Manager, Standards Development Division
OCIO/SDMG
571-272-9004


-----Original Message-----
From: Len [mailto:cbullard@hiwaay.net]
Sent: Saturday, January 10, 2009 11:49 AM
To: 'Costello, Roger L.'; xml-dev@lists.xml.org
Subject: RE: [xml-dev] What makes a data component good for
standardizing?

The user anticipates you.

It is a negotiation.

len

-----Original Message-----
From: Costello, Roger L. [mailto:costello@mitre.org]
Sent: Saturday, January 10, 2009 8:34 AM
To: 'xml-dev@lists.xml.org'
Subject: [xml-dev] What makes a data component good for standardizing?


Hi Folks,

Suppose you set out to create some standard data components. Your goal
is to
improve interoperability by creating standardized data components.
Particularly, you want these standardized data components to improve
interoperability between systems that weren't a priori coded to
understand
each other's data exchange format (i.e. you want to improve
interoperability
with the "unanticipated user").

What makes one data component good, and another bad?

(By "good" I mean the data component would in fact help improve
interoperability with the unanticipated user. By "bad" I mean the data
component would do little, if anything, to improving interoperability
with
the unanticipated user.)

I'll share my initial thoughts. I'd like your feedback on my initial
thoughts, and I'd also like to hear your thoughts.

Note: by "data component" I mean a chunk of markup that can be reused in
multiple XML vocabularies.


MY INITIAL THOUGHTS

I think that some data components would be good to standardize, while
others
would not be useful.

I'll start with two examples of data components would be good to
standardize.

Think about a postal address. It would be a good data component to
standardize. It's a useful data component even if I don't understand the
context in which it's being used.

      For example, suppose some nuclear physicist unexpectedly sends
      me a document containing data that I have no clue
      what it means, but embedded in it is a postal address.
      I may not be able to process all that data about
      subatomic particles (quarks, neutrinos, etc), but I can
      pluck out the postal address and store it in my address book.

That's interoperability between unanticipated users, albeit limited.

Another example of a useful data component is a business card (vcard).
Again, that's a useful data component that I can immediately utilize,
even
if I have no clue what the rest of the document is talking about.

These data components are useful independent of their context. I can use
the
data components even if I can't use all the stuff that they reside in.

Now I'll give an example of a data component which I think would not be
useful to standardize.

Both postal address and vcard gives a person's name (along with other
data).
Suppose I decide that I want data components with finer granularity than
postal address or vcard. Would "person name" make a good component for
standardizing?

I think not. A person's name would not be useful independent of context.


      For example, the same nuclear physicist above
      sends me the same document but containing the
      standardized PersonName data component, about
      a person named "Jim Brown.
      I am PersonName-aware so I am able to pluck out that
      Jim Brown information, even though I have
      no clue what the rest of the document says.
      Have I gained anything? No. It could be Jim Brown
      the ex-football player or some other person by that name.
      To make sense of the data component I need to
      understand its context.

I propose these two metrics for evaluating the usefulness of data
component:

    1. The data component must be standardized
       and broadly adopted (see below).
    2. If I can meaningfully use the data component
       without understanding any of the context in
       which it resides then it is a good data
       component. If I must understand its context
       then it is a bad data component.

Standardizing is good. It enables two parties to understand each other,
i.e., interoperate.

But standardization is not enough. I want more than interoperability
between
two parties that have a priori agreed to a data interchange format. I
want
interoperability between two parties that haven't a priori agreed to a
data
interchange format. I want interoperability between unanticipated
parties.

So the key is to not only standardize, but standardize the right things.


SUMMARY

We would go a long way toward advancing interoperability of
unanticipated
systems if we focused on creating standardized components that are
useful
independent of context.

What do you think?

/Roger
_______________________________________________________________________

XML-DEV is a publicly archived, unmoderated list hosted by OASIS
to support XML implementation and development. To minimize
spam in the archives, you must subscribe before posting.

[Un]Subscribe/change address: http://www.oasis-open.org/mlmanage/
Or unsubscribe: xml-dev-unsubscribe@lists.xml.org
subscribe: xml-dev-subscribe@lists.xml.org
List archive: http://lists.xml.org/archives/xml-dev/
List Guidelines: http://www.oasis-open.org/maillists/guidelines.php