[Home] [By Thread] [By Date] [Recent Entries]

  • To: 'Elliotte Rusty Harold' <elharo@m...>
  • Subject: RE: The triples datamodel -- was Re: Semantic Web permathread, iteration n+1
  • From: "Bullard, Claude L (Len)" <len.bullard@i...>
  • Date: Tue, 8 Jun 2004 11:37:36 -0500
  • Cc: XML Developer List <xml-dev@l...>

If the validation is Draconian, that's true.

What if the process is to allow the message to pass but to route 
the schema output into an audit file?  Nothing is held up process 
wise, but ontological commitment (for lack of a better term) 
is tested and variations are recorded to fold back at a later time?

As I said, eagerness and the message topic are everything.

len


From: Elliotte Rusty Harold [mailto:elharo@m...]

At 9:15 AM -0500 6/8/04, Bullard, Claude L (Len) wrote:
>That's good.  Except the bit about costs going up.  Why
>would they?

Costs go up because you need more human intervention with a 
conservative schema than a liberal one. If you require every XML 
document that adds extra information you don't care about and haven't 
seen before to be inspected by a human, that's more documents to look 
at and therefore more human intervention; and humans are expensive 
compared to computers. Depending on the system maybe the rise in cost 
would not be prohibitive, and perhaps in some systems you could save 
more than the cost if the humans avoided mistakes the automated 
systems might make. Then again maybe not. As John Cowan alluded to, 
in financial systems the cost of delay can be huge, and a system that 
takes the extra time for human verification might well cost more than 
correcting the mistakes in a fully automatic system. A lot depends on 
the context.

Site Map | Privacy Policy | Terms of Use | Trademarks
Free Stylus Studio XML Training:
W3C Member