[Go-essp-tech] Server side processing

[stephen.pascoe at stfc.ac.uk]

Hi all,
 
After the meeting in Hamburg I had some thoughts about Steve Hankin's
talk on server-side processing.
 
As I said at the meeting I believe encoding expressions into a URL is
going to become increasingly restrictive as the expression syntax gets
more complex.  Apart from the sheer length of URLs there will be the
issue of grammar and encoding.  For instance you if want to refer to
other OPeNDAP resources you would have to embed one URL in another,
requiring quoting, escaping, etc.  This could get very complex and run
into problems of multiply escaped characters.  I've seen it happen in
other web services.
 
If we want to build on tools and standards that our community already
use I would recommend extending NcML to include processing tags and
developing services that accept NcML as input (for instance as the
body of a HTTP POST or PUT).  You could imagine a very RESTful system
where a server allows new virtual datasets to be created by PUTing
NcML.  The result of such a PUT could either be NcML on the server
that constructs the virtual dataset on demand or the NcML could be
"executed" on the server to create a NetCDF.
 
I only have a shallow knowledge of NcML so I don't know to what extent
this is being done already.  In TDS has NcML been used to describe
virtual
datasets other than aggregations?
 
Cheers,
Stephen.
 
---
Stephen Pascoe  +44 (0)1235 445980
British Atmospheric Data Centre
Rutherford Appleton Laboratory
 

-- 
Scanned by iCritical.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/go-essp-tech/attachments/20091009/f02455f8/attachment.html