Peter Baumann wrote:
John Caron wrote:
Martin Daly wrote:
Sorry, I didn't explain myself very well.  What I meant was, if the 
full extents of the coverage are in the range [0,180] and [0,90], 
then the client has no way of knowing what the origin is.  That is, 
the client could detect longitudes > 180 and latitudes > 90 and 
deduce that the server is using the [0,360] and [0,180] ranges.  But 
that hueristic breaks down for smaller extents.
 
Like I said, the extents ranges should probably respect the CRS 
extents and origin.  That way there is no guesswork involved.
 
Just trusting the results might not be enough.  For example, our 
client allows the user to clip the coverage request to the extents of 
the current view.  We need the current view extents and the coverage 
extents to either be in the same domain, or to be able to know the 
domains.
 
Regards,
Martin
This reminds me of a question Ive been meaning to ask.
Suppose the client asks for an area that is bigger than the actual 
data area. I assume that we should return the intersection?
What if the client asks for an area that doesnt intersect? Is that an 
illegal request? Do we return an empty file?
my spontaneous reaction: both is an error. The spatial extent is 
advertised, and the client must keep with that. Otherwise we more and 
more deviate from the correlation between what is being asked and what 
is being returned.
I don't feel happy with an empty file - if this indicates a nonregular 
situation, it should be flagged as such = an error code should be returned.
Otherwise the client, in particular a non-leight-weight one,  may run 
into dozens of case distinctions before being able to really process 
what's been received. My personal opinion is that this would take away 
some elegance and clarity from WCS.
I know that WMS also allows a "far zoom out", which may be nice for 
viewing (well...some people may think so), but again: the primary 
mission of WCS IMHO is to serve _original_ data.
just my 2 cents...
The first should return the maximum extent advertised.  The second is an 
error.
just _my_ 2 cents...
gerry
Texas Mesonet