Roy has raised the issue (my paraphrase) of the form that a
URL should take to initiate a server-side computation.
Let me propose as a starting point the following approach,
which I had experimented with recently.
=Dennis Heimbigner
Unidata
Server-Side Expression Proposal
-------------------------------
A semantically nonsensical, but syntactically correct, example might
look something like this.
$x=f("string17",g(h(12))),f2($x,p[0:3:10])
The idea is that a URL would contain in its query section
(the part after the '?') a set of comma separated expressions.
An expression would be either
1. a function invocation, or
2. an assignment of a function invocation to
a single-assignment variable.
A function invocation is your typical one of the form
f(arg1,arg2,...,argn)
where the arguments can be any of the following.
1. another, nested, function invocation
2. a constant (string, number, etc)
3. a dataset
4. a variable.
The value of a function is a dataset (see below) so that
functions can be nested/composed.
The term "dataset" is intended to cover all of the following
possibilities.
1. whole dataset (e.g. a DAP Dataset or a netcdf dataset)
2. a variable within a dataset (variable here is not be confused
with a single-assignment variable.
3. a range expression applied to a variable within a dataset
p[0:1:30], for example.
The term "single-assignment" means that a variable can have
a value assigned to it exactly once. The reason for using
single-assignment is that allows for the representation of
any possible data-flow without introducting the notion of
mutable state.
The above is just a sketch, and there are many details
I am ignoring (such as typing and namespaces).
=Dennis Heimbigner
Unidata
Roy Mendelssohn wrote:
Hi All:
Ah, as usual I get to be the wet blanket. I am not as excited about WPS, or
for that matter most OGC standards, as others, because I find since OGC tries
to answer the general question, the resultant services have to deal with the
most complex case, therefore the 80% of uses cases that aren't that complex are
burdened with overly complex services, that are difficult for the average user,
and require constant update of software, and software that is hard to embed
into existing applications.
Most of what a lot of us do are essentially read-only services. For these, I
much prefer the basic approach presently used by F-TDS and GDS, though there
can be discussion of how best to connote an expression, and what the syntax
should be. Even more, as has been noted, WPS does not solve the key question
of the syntax of the expressions.
So I know this will start a lot of flames, but I just look at the service that
we developed and use, ERDDAP. All sorts of people, who only know a little
programming, successful use ERRDAP in R, Matlab, shell scripts, python, java
etc., and it is easy to use from within applications, because any environment
that can send a URL and receive a file can use it. Moreover, since only
methods internal to the applications are used, the scripts or programs or
whatever do not break as both the application and operating environment change.
Our experience over many years of serving data is that in most cases, simple
and fast are to be preferred. Clearly people can come up with counter cases
that are not covered, and I am all for developing services for those cases, but
not for burdening simpler cases with all that baggage.
My $0.02.
-Roy
On Jun 19, 2012, at 12:14 AM, <stephen.pascoe@xxxxxxxxxx> wrote:
We are working on wrapping CDO operators in our WPS implementation COWS-WPS. This work is being done as part of the IS-ENES an ExArch projects with with DKRZ, KNMI and others.
WPS doesn't define your functions for you but gives you a protocol and framework for asynchronous server-side tasks.
http://cows.ceda.ac.uk/cows_wps/intro.html
--
Stephen Pascoe from iPhone
On 18 Jun 2012, at 23:07, "John Cartwright" <john.c.cartwright@xxxxxxxx> wrote:
Could the the OGC's Web Processing Service spec
(http://en.wikipedia.org/wiki/Web_Processing_Service) serve that
function?
--john
On Mon, Jun 18, 2012 at 3:56 PM, Roy Mendelssohn
<roy.mendelssohn@xxxxxxxx> wrote:
But for a useful service, the form and syntax of the URL should be independent of the
mechanism that does the server side calculations (which rules out SWAMP). So for
example, both Grads an F-TDS use the same format in the URL to say that "this is an
expression", but the expressions themselves are platform specific. That is not the
way to get overarching services.
Instead, we need agreement on how in the URL request we signal a server-side
function, the syntax of that function (independent of the engine underneath)
and a few simple functions to start (say simple data transformations,
differencing and averaging on a dimension(s)). Then the server back-end can
parse the request and use Ferret or Grads or NCO or Python or whatever is
desired, and like with any good service, the back end could change without
having any affect on the URL or the user.
I know Matthew Arrott at least used to like the approach in Chapter 12 of "Python
Scripting for Computation Science". But a lot of that is the engine underneath. I
am more interested in the form of the URL. Get some agreement on that, and some real
implementation could proceed.
-Roy
On Jun 18, 2012, at 2:37 PM, Russ Rew wrote:
Jeff,
However, we have to keep in mind performance ramifications. It still takes
a long time to move gigabytes of data across a network. This brings up the
importance of moving the computation to the data, instead of moving the
data to the computation. For some data sets and many use cases remote
access to data works very well so things like brokering are tractable.
However, for *big* data sets (e.g., climate model output) we need to come
up with richer mechanisms (like the NCO on local data) to bring computation
to the data.
See Daniel Wang's SWAMP (the Script Workflow Analysis for
MultiProcessing), built on top of NCO:
https://code.google.com/p/swamp/
--Russ
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
**********************
"The contents of this message do not reflect any position of the U.S. Government or
NOAA."
**********************
Roy Mendelssohn
Supervisory Operations Research Analyst
NOAA/NMFS
Environmental Research Division
Southwest Fisheries Science Center
1352 Lighthouse Avenue
Pacific Grove, CA 93950-2097
e-mail: Roy.Mendelssohn@xxxxxxxx (Note new e-mail address)
voice: (831)-648-9029
fax: (831)-648-8440
www: http://www.pfeg.noaa.gov/
"Old age and treachery will overcome youth and skill."
"From those who have been given much, much will be expected"
"the arc of the moral universe is long, but it bends toward justice" -MLK Jr.
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit: http://www.unidata.ucar.edu/mailing_lists/
--
Scanned by iCritical.
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit: http://www.unidata.ucar.edu/mailing_lists/
**********************
"The contents of this message do not reflect any position of the U.S. Government or
NOAA."
**********************
Roy Mendelssohn
Supervisory Operations Research Analyst
NOAA/NMFS
Environmental Research Division
Southwest Fisheries Science Center
1352 Lighthouse Avenue
Pacific Grove, CA 93950-2097
e-mail: Roy.Mendelssohn@xxxxxxxx (Note new e-mail address)
voice: (831)-648-9029
fax: (831)-648-8440
www: http://www.pfeg.noaa.gov/
"Old age and treachery will overcome youth and skill."
"From those who have been given much, much will be expected"
"the arc of the moral universe is long, but it bends toward justice" -MLK Jr.
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit: http://www.unidata.ucar.edu/mailing_lists/
