> -----Original Message-----
> From: Nathan Scott [mailto:nathans@xxxxxxxxxx]
> Sent: Friday, 17 April 2015 8:37 AM
> To: Frank Ch. Eigler; David Smith; Ken McDonell
> ...
> Is there any reason not to tackle this problem, Frank? There's a possible
> solution proposed reusing existing code, & other PMDAs that will benefit -
> seems like a no-brainer that we should Just Do It.
I suspect this discussion needs to fork, as there appears to be 2 related, but
different, "topics"
1. seeing if the pmdaCache code could be reused to provide a more general,
optionally persistent, name <--> id mapping cache that could be used instance
domain management and other tasks that might help the PMDA implementer .... on
this topic I suspect there is probably not much more to be said, I'm with the
"Just Do It" camp here.
2. trying to properly document what is expected of a PMDA implementer in terms
of data invariance over time ... we have lots of implicit assumptions here, and
fleshing out all of the issues would probably help people new to PMDA
development and at the same time give us a framework to judge how well client
tools might operate if the data from a PMDA was to change over time (I'm using
"data" here to mean not the values for metrics, but all of the metadata, the
metric names, their PMIDs, ...).
|