Standardising units and formats (and ref frames?) in transmission

[Douglas Tody]

On Fri, 15 May 2009, Patrick Dowler wrote:

> On Friday 15 May 2009 11:10:19 Alberto Micol wrote:
> > I really hope to hear more (public) voices on this (old) idea of
> > standardising units and formats
> > --if not even reference frames-- for the most common and relevant things
> > that travels onto the wire
> > (not what is presented to the Users).
> ...
> > The approach here proposed (again) is just a transport mechanism, an
> > infrastractural standard,
> > nothing for humans to see, to interpret, or to be affectively attached to.
> 
> 
> It seems rather inconsistent to say that astronomers will build their own
> tools and at the same time these standardised units and formats will not be
> "seen" by the users.

It is the users of the applications built by astronomer/programmers (who
are the ones using our services) who do not see what goes over the wire.

In any case, there has to be a middle ground here as usual.
We standardize primarily where we add things, such as standard
protocols, or mediation to a standard model.  If we are exposing native
data then we want to be careful about changing the content, but adding
a few table fields, or some standard metadata in a DAL query, is ok.

In general the guideline for this sort of thing (not just in astronomy)
is to limit the choices for units etc. so far as possible in the wire
protocol, with fan-out to all the possibilies at either end (data
source and client application).  If we try to support all options at
all levels we will probably be less successful than, e.g., CORBA.

> I agree that users need to be able to write their own task-specific tools -
> that would be great and a sure sign that the VO works! My feeling is this
> will happen when someone does enough of the (hard) work and content is
> uniform enough for users to succeed. I don't think we can mandate that data
> providers do that work, but we can encourage them, make it easy for them,
> and maybe in some cases someone else will do it for them (eg. make a service
> with the same content but with extra computed columns in standard
> frames/systems/units/whatever... value-added web 2.something goodness).
> 
> Or they could all use STIL and Mark will have to do all the hard work..
> sounds like plan B :-)
> 
> Seriously, I think first we need to make the content available in a more
> technically uniform fashion (IVOA service standards). Then we can see about
> more uniform content.

Making the standards robust is certainly critical (and we are not even
close yet despite trying to minimize complexity!).  The observatories
and projects would probably appreciate it though if we provided some
guidelines for common things which are fairly generic and standard,
such as exposing index tables and catalogs.

 	- Doug