Re: [Cbor] Money amounts

[Carsten Bormann <[e-mail]>]

On Feb 27, 2019, at 13:41, Changaco <changaco=[e-mail]> wrote:
> 
> On Mon, 25 Feb 2019 11:06:57 +0000 Richter, Jörg wrote:
>> And I'm not a big fan of encoding additional information in the number of 
>> decimal places.  So "EUR10.00" should be transmitted as "EUR10", or ["EUR",10].
>> But for other decimal fractions you are right that the string might be shorter.
> 
> I think it's best to be lossless by default. If an encoder receives a `10.00`, for example a `Decimal('10.00')` object in Python, then it should encode the trailing zeroes unless it has been explicitly configured to drop them.

But that presumes there is a semantic difference between “EUR10” and “EUR10.00”.  Is there is there not?

Assuming there is no semantic difference, we could still make it so that an encoder can output both, and a decoder is required to be able to process both, presumably with an identical result.  Or we could use the concept of “preferred encoding” with that, with making the shortest encoding the preferred one, but not placing the onus on the encoder to always generate preferred encoding (outside instances of canonicalization), but continuing to place the onus on the decoder to handle both, but *not* placing the onus on it to keep the distinction available to the application.

Interoperably handling equivalent representations of the same information has turned out to be more laden with consequences than one would initially think…

If there *is* a semantic difference, it should be clearly explained, or even better, made more explicit in the encoding.

Grüße, Carsten