[QUANTITY] Choosing default accuracy/units values

[Patrick Dowler]

I agree that default units and default accuracy/error should follow the same
rules. However, to be successful the VO really needs to get people (force them 
if necessary) to not be lazy, so I would prefer making specification of units 
and accuracy mandatory (in the schema) and have some special values that can 
be easily used, like <unitless> and <exact> and <uknownUnits> and 
<unknownAccuracy>. There should be no default value.

Yes, this does potentially increase the size of the serialised form, but the 
benefits outweigh the cost, IMO. And sure, those unknown tags are especially 
large, but they should be rare (who wants data with those tags in there?) and 
providers of such tags/data will be discouraged by their usage stats :-)

my 2c,

Pat

PS-If serialised size becomes an issue, it only matters for large docs and in 
those cases there are other ways to optimise. For example, if a doc has many 
Qs with only 2-3 different unique unit strings, one could form 2-3 groups and 
factor the common parts up one level to the group. The group would contain 
the Qs that share some things and the shared things. This kind of thing would 
appear in the schema and in the readers/writers, but wouldn't have to be in 
any other software (unless such a concept is generally useful... for now I'd 
leave it out, then I'd consider adding it and hiding it, then I'd consider 
making and making it visible - ie not removing it during parsing).

On Monday 17 May 2004 10:05, Brian Thomas wrote:
> On Monday 17 May 2004 11:57 am, Jonathan McDowell wrote:
> > The counter-argument is that <unitless> is really common, especially
> > for integer and string data. What do other people prefer?
>
> 	There are good "for" and "against" reasons one can offer for
> either
> 	"unitless" or "no-units-defined" as a default, so I think this
> is more of
> 	a religious issue than anything else. For me, I'm a "unitless"
> man.
>
> 	Where this issue can be thought of more "rationally" is that the
> choice
> 	of "unitless" or "no-units-defined" as a default should be
> consistent (ideally)
> 	with the same choice for "accuracy" e.g. "ExactAccuracy" or
> "NoAccuracyDefined"
> 	(the latter being the "lazy" case). Thus, a choice of "unitless"
> means "exact"
> 	for the default cases as the same rationale applies (much
> integer/string data
> 	are constant/definition type so that means its implicitly
> unitless/exact value).
> 	Conversely, "many people are lazy about what the real
> units/accuracy is and just
> 	leave it out" would have the consistent choice of
> "no-units-defined"/"no-accuracy-defined"
> 	as the default choices.
>
> 	=b.t.

-- 
Patrick Dowler
Tel/Tél: (250) 363-6914                  | fax/télécopieur: (250) 363-0045
Canadian Astronomy Data Centre   | Centre canadien de donnees astronomiques
National Research Council Canada | Conseil national de recherches Canada
Government of Canada                  | Gouvernement du Canada
5071 West Saanich Road               | 5071, chemin West Saanich
Victoria, BC                                  | Victoria (C.-B.)