I think in general the overhead in maintaining current information
about old registered values is too high, and that it *is* worth time
thinking about how we could lower the overhead for registry
There are a number of related issues raised about various registered
values, including MIME type, charset, and URI schemes.
Ideally a registry is a place where a new implementor can go to
discover both the theory and current practice for use of registered
values on the internet. I think the current processes cope OK with
theory (although the overhead of updating the registry when there is
a new spec is high, it might be acceptable) but not with practice
(where implementation and deployment sometimes is in advance of, or
divergent from, the formal specs).
The situation is more acute in areas where protocols and formats are
undergoing rapid development.
So I agree that writing a document marking 'afs'
as 'obsolete' is make-work and not-worth anyone's time, but how
could we make it easier (light-weight annotation) without subjecting
ourselves to DOS of unreliable annotation?
I'm forwarding this message from apps-discuss list here since
the discussion is also appropriate here.