[QUANTITY] Why quantities always have errors

[Martin Hill]

Brian Thomas wrote:

> On Tuesday 18 November 2003 11:29 am, Martin and Brian wrote:
> 
>>>I think you are making the point that where there is no error with a
>>>value, we should make it clear whether it is because we don't know the
>>>error, or that there shouldn't be one.  I believe this discrimination
>>>sould be in given in the 'type', as (1) it shows much more readily when
>>>an error is relevent or not, and (2) it 'forces' the user of the type to
>>>consider the error, rather than habitually setting it to 'None'.
>>
>>        Yes, agree, this is my point of view, and why I want "error" (or
>>        actually, I'd call it "accuracy") on all quantities.
> 
> 	Opps, I need to modify my agreement here. I think this makes sense
> 	(puting/connecting value with data type) when you have something like 
> 	"string" data type, which is always of a particular accuracy ("exact, no error"). 
> 	Not sure thats right for the numbers or not. Recall that some errors are "universial" 
> 	and apply to all values in the quantity (such as for a systematic error) but others
> 	apply directly on a value by value basis. Since the data type is a "universal"
> 	itself (all values have the same type), you have to be carefull for specifiying 
> 	accuracy there when numbers are involved.

Gah I don't understand this, probably because we're using the same words 
for different things.  To me type=class/interface (so Flux, Co-ordinate, 
even Quantity if there is such a thing, are all types); are you using it 
for 'primitive types'?  Why is a data type universal?  is a 'value' not 
a quantity?

I will shut up soon, honest.

-- 
Software Engineer
AstroGrid @ ROE
Tel: +44 7901 55 24 66
www.astrogrid.org