[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [emergency] Naming Conventions
Another path through this mess is what CDISC has done. The FDA published a set of rules for submitting drug information electronically so a pharma could get its hot drug approved. CDISC [clinical data information standard consortium] was formed from vendors, pharma companies and others to jointly develop an implementation standard using xml. They have defined the tree structure with controlled names. NCI used common data elements (CDEs that I spoke about) to handle the data element meta data, the instances of data within a specific node of the tree. The fact that there were real world objects [forms and data on them] enabled this project to move forward pretty quickly. I have been involved in too many data element definition project for DOD in the past that tried to do it all and failed miserably because real world apps and their users were not involved.
John
-----Original Message-----
Got it, Rex. The usual meta-meta problem: do a data design a la abstract interface or create a vocabulary. I like the VRML solution to that: do both with the encodings and vocabularies as optional features. It's more work but one can nail down the semantics and give the user community a chance to converge at the level of efficiency most appropriate to the local contracting situation. When volunteer groups sit down to create specs and standards, they should keep in mind the problem for the actual industry of changing horses. We're glad to see the work being done by OASIS in public safety, but until it shows up in an RFP, it's just nice-to-have, Sounds Good Maybe Later, kind of work. DOJ can't drive it down from the top without funding it too. And technical standards that don't account for medium constraints don't have a snowball's chance anyway. Thus, we use the RFP to guide us and cost accordingly.
I note reading through the JusticeXML dictionary that they have wisely left open most aspects of the data definition (eg, loose types, loose occurrence constraints, etc) and stayed mainly to a labeled tree. Much easier to live with that way.
len
|
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]