time-stamp conventions

NetCDF Group,

The Energy Sciences Department (ESD) at Pacific Northwest Laboratory 
studies energy use in equipment, buildings, and systems.  These studies
largely depend on end-use metered data.  ESD analyzes metered 
data in time-series or database format using various software packages 
on multiple platforms.

ESD is implementing NetCDF for our metered-data storage format.  We chose
NetCDF because of the ability to document our data in the file and
because of the ability to append the file in the unlimited (time)
dimension.

We feel that the 8 byte time stamp (integer value is day, decimal portion
is fraction of day) is the most flexible because it allows a trade-off 
between precision (small time-steps) and size (maximum days accomodated).
But we do not wish to store an 8-byte time-stamp in the same variable as
4- or 2-byte values.  Is it in line with NetCDF philosophy/user practice 
to have a separate time variable that refers to a data variable?

If so, how are record variables treated with inconsistent time-stamps?

Has anyone tried to standardize time formats for data with regular time-
steps, data with irregular time-steps, and data that is recorded only
occasionally?

Thanks,

Andrea Wood
Energy Sciences Department
Pacific Northwest Laboratory
(509) 375-3934
ag_wood@xxxxxxx