WD-AccessData-1.0-20140312

[Petr Skoda]

> I just have to add my 2-cents worth and say that I think this is wrong.
> The protocol should be as simple as possible, and adding bandpass names is
> not.

I think you did not see the previous discussion - WHY the bandpass names 
IS important:

1) It is the primary information written in FITS header - the provider of 
the original data (e.g. camera acquisition system) always uses some 
simplified names, lists ....

2) The names of particular filters MAY NOT be mapped precisely to the 
filter names - it is spcific to every project and there is not unique 
solution how to map it - but of course the filter profile service 
used in VOSA etc. tries to identify the most commonly used.

3) The scientist know MUCH BETTER what kind of filter the particular 
survey (say resource) uses then the VO data provider can describe in 
metadata (and if he wants to be precise it requires a lot arguing with the 
expert in e.g. photometry - moreover some filters have discontinuity (they 
absorb) in certain regions so the only physical identification of it is 
the plot of transmissivity.

4) It may be even worse - the instruments with many filters may such that 
are not described well by simple names. it may be e.g. Hbeta filter 
centered on some wavelength coresponding to the given redshift of this 
line. Some documentation of filters mentions only central wavelength not 
min and max range - Some state central wavelength and FWHM or half-width 
...


So it is almost impossible for the (usually informatician with basic 
astronomical background) VO data publisher to find the exact mapping 
required here.
And how will you distinguish two different images the filters of which 
have the similar wavelegth range or central wavelength - e.g. Johnson V 
and Stromgren y .....?

BTW - the polarisation states are as well completely wrong and they should
d not be listed as Q,U,V,I as none of this value is obtained separately 
and you cannot use it for query - but you may want to extract them in 
Accessdata - see the presntation of F. Paletou in Heidelberg.

But it is more practical to refer to obseravtion with certain orientation 
of filter (half-wave plate) as e.g. U and to the one without polariser as 
I....

>
> If you want a user interface or toolkit that layers bandpass-based queries
> on top of the basic protocol, that is fine.  User sees bandpass name = V,
> protocol sees a wavelength range.  But everything you add onto the
> protocol is another feature the data provider has to implement.
He has to implement much more complicated translation to wavelength ranges 
(very artificial and imprecise ) just to satisfy the characterization.
But here it may be considered as the rough orientation value and it is 
legitimate use it in (very rough) discovery query to get at least 
something.
But if the scientist need EXACTLY what he wants, there is no way how to do 
it in a unique way.