I do not understand where David's implication
was drawn from. Certainly having a schemaLocation
for a schema does not necessarily mean that there
are no local caches. There are also nowadays some high
availability, fault tolerant, redundant, failover, load balancing
systems.
This means that though there is one URL,
there need not be, in any pejorative sense, "a single
point of failure". A system backup is also possible.
DNS IP pools for the host, parallel live systems
running in sealed tunnels in Colorado, etc.
What does any of this have to do with updating
a schema as a work-around for an interop. problem
arising from a discrepancy between normative words in a
specification and the normative schema of that
specification.
I guess I must be missing something.