xml-dev - Re: Namespace and Versioning

[james anderson <james.anderson@mecomnet.de>]

"Bullard, Claude L (Len)" wrote:
> 
> ... What I was looking for initially
> in this thread is some comments on the practices
> people are using to see if any general cases
> emerge. ...
> 

to that point i can describe the method we implemented to handle
versions of encoding specs from the Kassenaerztlichebundesvereinigung
(KBV). (they're the group in germany which sets standards on electronic
data interchange within the health industry.) they have the habit of 
every several years bringing out a new standard. often enough, the new
standard ascribes different meanings to the various terms which were
present in the previous standard. there are also cases in which the same
term has different meanings in distinct contemporaneous standards.

the implementation domain was lisp, so the name-management was built on
the package system. each namespace was modeled as a package. each name
was modeled as a symbol. each symbol was associated with the appropriate
processing method.

the package system provides mechanisms to control identity relationships
between names. this both in terms of inheritance relations between
packages and in terms of specific assertions regarding individual symbols.

each KBV version was associated with a package. the package names, and
thus the namespace names were, in fact, chosen to reflect the relations
among versions, but that was immaterial to the implementation. the
package relations were arranged such that terms for which the meaning
changed were associated with distinct symbols in the respective
version's packages. terms for which the meaning had not changed were
carried over from the older version's package to the newer version's by
"inheriting" the respective symbol from one package to the next.

it was much easier to do than to describe. the moral is, you need the
means to control relations between namespaces; the namespace names
themselves may well serve documentary ends, but are, in fact, secondary.