Re: Constraints and Functional Dependencies

On 2007-02-28, paul c wrote:

> What is probalistic about it? I thought I could convert the same TIFF
> to a JPEG twice and get the same JPEG both times.

Still, the end user might be having two different days. Unlike the DBMS, the end user won't be repeating operations exactly. Hence, what the end use wants varies from day to day. The "equivalence" between two images that he wants varies all the time, and can only be modelled probabilistically.

Secondly, by definition you cannot exactly convert the other way around. Once you have a JFIF, at best you have an entire, countably infinite class of decompressed TIFF's you can convert it into. (Or was it the other way around? Either way, it doesn't affect the argument.) Many of those resultant TIFF's will also compress into JFIF's which aren't equal to the original one, or necessarily in the equivalence class naturally induced by the preimage of the resultant instance of JFIF.

As I said, we're not talking about transitive relations, here. I'm not even aware of an a pair of a compressor and a decompressor that are generalized inverses; that means it's difficult to even get reflexivity to hold, here.

-- 
Sampo Syreeni, aka decoy - mailto:decoy_at_iki.fi, tel:+358-50-5756111
student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
openpgp: 050985C2/025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2