On 3/22/2011 12:56 PM, Steve Hankin wrote:
On 3/22/2011 5:40 AM, John Caron wrote:
On 3/21/2011 11:55 AM, Karl Taylor wrote:
Dear all,
I haven't had time to follow all the discussion in detail, but in
general I think CF should not add additional complexity unless the
current way of encoding time is incomplete. As far as I know the
encoding is indeed complete and given correct specification of the
units (which include basetime) and a calendar, the calendar
date/time can be calculated. This indeed requires a smart library,
but I think that Bob Drach's CDMS correctly performs such a
calculation.
I'll try to go back and read the arguments, but I think I agree with
most of what Steve Hankin has said.
Best regards,
Karl
Hi Karl:
There are 2 things incomplete from my POV:
1) CF specifies calendars, but theres no reference library that
implements them. If CDMS does so then perhaps we can leverage that.
2) We need to be able to express/manipulate calendar dates better
that we can now with udunits.
Good summary.
I guess I could just say that im disatisfied with udunits as a
reference library for calendar time units. The grammar allows months
and years as time intervals when that doesnt make sense. There are
some other things that are unclear (ill post one in a second).
Anyway, I think the reliance that CF has on udunits is, um, suboptimal.
No disagreement that udunits is "suboptimal" -- incomplete formatting
functions and having a couple of small inconsistencies (e.g. "months"
as a unit). Arguably, though, it is the best currently available
implementation and data model for the CF problem space. So shouldn't
we simply be talking about improving udunits? While there is no
current reference library, the required concepts are already very well
tested in code -- decades of successful practice in multiple
applications, including Ingrid, CDAT, Ferret, GrADS (presumably) and
others). What is missing is the human resource to formulate a
community specification and implement it. (Are you volunteering?)
sure - you do mean implement in Java, dont you ;^)
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
