Hello Jon, Roy, John, Ben, and all,
it's exciting to follow this discussion here. For documentation and
consolidation, would it be possible to somehow cast all these valuable
facets into the use case page @ GALEON?
thanks a lot,
Peter
Jon Blower wrote:
Hi Roy,
You are quite right of course that subsetting servers are important to
scientists (I did say that I was generalizing!) I was trying to
highlight that "dumb" servers have their advantages because scientists
can trust that they haven't manipulated the data. OPeNDAP I guess
falls somewhere between the "dumb" (e.g. Apache) and "smart" servers
(e.g. WCS) because it does basic subsetting in index space and hence
its output is predictable - it behaves exactly as if the users have
done the subsetting locally with no other manipulation and hence is
pretty trustworthy. I don't want to suggest for a moment that
server-side subsetting isn't important (we use OPeNDAP every day too)
but scientists tell me all the time that "we just need the data".
Again, I just wanted to say that "how" something is done is just as
important to use case analysis as "what" is done.
Regarding John G's point about sensors - I'm afraid I don't know
anything about sensors but I guess all data is processed somewhat
before publication: even model results are often translated to a
different format (e.g. Met Office PP format to NetCDF). I think the
point is that scientists often feel most comfortable getting data from
as close to the original source as possible, as the data providers
intended it.
Cheers, Jon
