I'd prefer a configurable setting rather than a custom URL parameter. I try to
avoid extending specifications if at all possible. No existing clients will
know about the additional parameter, and some datasets won't benefit from
start/stop/step. On the downside, the GetCap could list all of the times for
some datasets, and use start/stop/step for some others. At least with a URL
parameter, it would be consistent.
I took a simple stab at #1 a few weeks ago. The rounding isn't triggered by a
flag in the GetCapabilities request, it is instead enabled by using a checkbox
in the admin panel on each dataset
It finds the smallest interval between all timesteps in the dataset and then
assumes that this interval is consistent throughout the dataset. It steps
through all of the timesteps and if the interval between two adjacent timesteps
is greater than the smallest found interval, it ends the current
"start/stop/step" and starts another.
>From what I've seen, it works, but I haven't tested it at all.
https://github.com/asascience/ncWMS/commit/9e2925fc607a05d6484299e017db0180a2200fa4
---------
Kyle Wilcox, Engineer
Applied Science Associates
55 Village Square Drive
South Kingstown, RI 02879
p: (401) 789-6224
e: kwilcox@xxxxxxxxxxxxxx
-----Original Message-----
From: thredds-bounces@xxxxxxxxxxxxxxxx
[mailto:thredds-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Jon Blower
Sent: Tuesday, January 04, 2011 5:54 AM
To: thredds@xxxxxxxxxxxxxxxx
Subject: Re: [thredds] Simple fix => much smaller TDS WMS GetCapabilities size
(for model output)
Hi all,
I can certainly see that there is a problem that needs to be addressed here
(explicit lists of all individual TIME values cause the Capabilities document
to blow up). There are actually two approaches to this, which could be used
individually or in combination:
1) Use the syntax start/stop/period, potentially multiple times, to define the
TIME values instead of listing them explicitly.
2) Use Layer inheritance properties to define the time dimension once only, if
the same time axis is shared by all the children of a parent layer:
<Layer>
<Title>My Model Output</Title>
<!-- Non-displayable parent Layer -->
<Dimension name="time"> ... values ... </Dimension>
<Layer>
<Title>sea_water_temperature</Title>
<Name>TMP</Name>
<!-- Inherits time axis from parent -->
</Layer>
<!-- More child layers -->
<!-- Children can override the time axis if theirs is different for some
reason -->
</Layer>
The most concise possible Capabilities doc would be achieved by combining both
approaches.
I feel that we should ensure that only those time values that are actually
present should appear in the Capabilities doc - I think things get a bit
confusing if the Capabilities doc advertises "missing" times (what would the
returned image from a missing time look like?). I also agree with Bob Simons
that the use of "nearest" values is dangerous, even though it's in the spec
(sorry Kyle) - the client can always perform the nearest-neighbour calculation
if this is required, given the server's advertised capabilities. (Feel free to
disagree with me of course!)
I can see two potential problems:
1. Solution 1 above is a bit tricky to implement in the general case, avoiding
the corner cases. (Solution 2 would actually be pretty easy to implement.)
2. As Ethan and Roy have pointed out, third-party client support for
multidimensional WMS is, er, generally not great. It's hard enough to find a
client that supports TIME at all, never mind all the possible syntaxes. I'm
torn on this - in one respect it's not our problem, but we don't want to cut
out portions of the user base.
So, after all this, I propose a solution:
1. Implement one or both measures above, ensuring that the Capabilities
document is accurate. This may involve being conservative. The default
Capabilities doc would be much smaller.
2. Allow clients to specify a URL parameter to GetCapabilities that triggers
the generation of a Capabilities document that *does* list all the time values
explicitly, allowing compatibility with some GIS clients. (Clients usually
require a URL to the Cap doc, which could include this non-standard URL
parameter. Or the parameter could be considered part of the "base URL").
Does anyone have any thoughts on this before I start an implementation? It's
tempting to implement the "layer inheritance" solution first since it's
easiest; I think this would be effective in TDS, where each Cap doc usually
represents a single model run, which will usually have a single time axis,
shared among all variables.
Happy New Year!
Jon
--
Dr Jon Blower
Technical Director, Reading e-Science Centre
Environmental Systems Science Centre
University of Reading, UK
Tel: +44 (0)118 378 5213
http://www.resc.reading.ac.uk
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
