[Units]

[Rob Seaman]

 From recent messages it doesn't appear that there is a single  
coherent proposal on the table yet.  Some notes, roughly from the  
point of view of VOEvent:

1) We need a master vocabulary of astronomical units.  (How prefixes  
are supported is an implementation detail.)  Everything else follows  
from the VO having a standard way of labeling quantities with their  
appropriate units for the full range of VO use cases from storage to  
protocols to user interfaces.

    1A) Before we can enforce any policy on conforming units, we need  
to recognize nonconforming usage.

2) Individual protocols will have different requirements for the  
quantities conveyed, and thus for the pertinent units.  Of necessity,  
each protocol will naturally correspond to a subset of the master  
vocabulary.

3) It may also be desired to construct a subset of units corresponding  
simultaneously to the complete list of protocols.  There are two  
issues with this:

    3A) The list of protocols is growing.
    3B) Should the "combined protocol subset" be the intersection or  
rather the union of those required for the individual protocols?

4) If the intersection, then a per-protocol mechanism is needed for  
special cases.

5) If the union, then per #3A there is a strong requirement for change  
management of the combined subset.

6) Individual VO *user interfaces* will also require different subsets  
of units.  Ditto for holdings associated with individual VO member  
archives.  In this case, my immediate reaction is that the master  
vocabulary is sufficient to cover them all.

    6A) Per #6, VOEvent will likely permit any broadly conforming  
unit.  Per #2, SEAP (the Simple Event Access Protocol) will likely  
permit only a subset.

7) VO protocols naturally require a restricted subset of the master  
list of units because the protocols each implement a limited range of  
methods. For instance, if SEAP permits a range query on celestial  
coordinates, then it must support the pertinent units.  If SEAP does  
not support a query on magnetic field strength, then it need NOT  
support tesla - or the more familiar non-SI gauss.

    7A) In any event, both tesla and gauss should appear in the master  
vocabulary because astronomers (those who the VO ultimately serves)  
may use both.

8) Suggestion: perhaps the master vocabulary should simply indicate a  
preferred subset approved for protocol interchange?  (Again, whether  
that subset attempts to be purely SI is a separate question.)

9) The individual protocols will then naturally each only use a sub- 
subset of the preferred units, because only some will be pertinent to  
each of their purposes.  If later the "Simple Whatever Protocol" adds  
a new method, it will naturally be restricted to the preferred subset,  
no change needed.

10) Both the master vocabulary and the preferred subset will be much  
more slowing varying than the rapidly evolving protocols, thus  
simplifying the requirements for change management.

Rob Seaman
NOAO