Hi Roy,
Thanks for your latest 3 posts.
They all make the point that hierarchical files
are widely used now. In this post you make the point that maybe
having conventions on how to use/interpret hierarchies would
be helpful, regardless of its desirability for CF.
Yes, part of our motivation is better conventions for hierarchies
independent of CF.
_And_ I do think there are
use cases ("ensemble", anyone?) that are desirable for CF
(which operates by consensus so let the users decide)
that could most effectively and concisely be implemented
with hierarchical files that utilize group-metadata inheritance.
So I posted the use-case for this to CF in case it wishes
by acclamation to help set such conventions. CR is an
important de facto standard with a track record of improving
dataset interoperability. Whether it wishes to weigh-in on
hierarchical "group-aware" structures remains to be seen.
cz
Le 18/09/2013 09:45, Roy Mendelssohn - NOAA Federal a ?crit :
> CF may well not be the proper place. I am not arguing that. All I am arguing is that the history of computer science shows that hierarchies are often beneficial compared to flat structures, whether it be the b-trees we use in our directories to structures in programming languages, while a lot of the discussion essentially has been "everything should always be flat, there is no need for hierarchies, search will take care of everything".
>
> I also know we deal with a lot of NASA data (files) and they are all over the map in their structure and how the hierarchy is used. So from an end users point of view, having some standard on how it is done and what we can expect would be a great help. There is a use case - whether that is compelling to CF is a different matter. And I think this is where Zender et al are coming from, because they see the same thing in the NASA data files.
>
> Another thing to consider is that netcdf4 now does have groups and hierarchical structures, and down the line, in particular I believe for in situ data, they make a lot of sense (I can give practical examples of this if people want). So there is some good rational to develop best practices for their use, before the use becomes common.
>
> My $0.02.
>
> -Roy
--
Charlie Zender, Earth System Sci. & Computer Sci.
University of California, Irvine 949-891-2429 )'(
Received on Thu Sep 19 2013 - 14:40:23 BST
