On 11/02/2014 02:55 PM, Ken McDonell wrote:
I should have been clearer in my earlier email ... any metrics that are only
in one archive will always be OK ... by "metadata differences" I mean a
metric that appears in more than one archive and has some difference in the
pmDesc data that describes that metric or the associated PMNS fragment that
maps a name to a PMID ... this means
- name
- PMID
- type
- indom (identifier of the instance domain)
- semantics
- units
The name to PMID and PMID to name(s) mapping we could handle (I owe Dave a
response to his email asking questions about this).
type differences we could handle, but I don't think it is worth the effort
and there are some combinations that cannot preserve precision.
indom and semantics differences are probably impossible to reconcile.
We could do units, but it is probably not worth the effort at this stage.
Different metadata between archives is sort of expected, as the examples in
Frank's mail illustrate.
I was referring to differences in metadata for the same metric.
In this context, my original brave assertion about this being OK most of the
time still stands, I believe.
Thanks Ken. This is essentially how I understood things.
Dave
|