[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: Converting a variety of data formats, containingvarious ki

• From: Peter Hunsberger <peter.hunsberger@gmail.com>
• To: Michael Kay <mike@saxonica.com>
• Date: Tue, 4 Oct 2016 15:06:12 -0500

Yeah, that makes sense, my wording is bad but that's why I qualified
it with "reducible".... It sort of the same point you made about
normalization and binary relational models: given enough triples you
can relate anything to anything. However, to put your point another
way, that doesn't mean that any given canonical triple model
represents all the same metadata of another canonical triple model
representing the same data... You can then, of course, create the
triple model that is the intersection of these two models, ad
infinitum.  There are likely some group theory (or category theory)
answers to what such models would look like when they are isopmorphic,
but all your models have to form groups of the same size for my
original statement to be true.  Guess I'm spending too much time
dreaming of a world where Odersky's Dotty (or similar) is the
foundation for everything...
Peter Hunsberger


On Tue, Oct 4, 2016 at 2:14 PM, Michael Kay <mike@saxonica.com> wrote:
>>
>> Without giving this much thought; you could probably prove that all
>> data modes that are "computationally  tractable" are reducible to a
>> model that is isomorphic to a triple store.
>
> Generally I don't think different data models are isomorphic except in trivial cases: you can't usually find a 1:1 mapping between their value spaces. If you try to find a function F that maps from the XML value space to the JSON value space, for example, you will typically find that if it's lossless (ie. represents every distinct XML value using a different JSON value, and therefore has an inverse F'), then it won't be the case that F' maps every distinct JSON value to a distinct XML value.
>
> Similarly if you map XML to a triple store, then the operation that reverses the mapping won't be useful as a general way of representing any triple store in XML.
>
> If such isomorphisms existed, it wouldn't be necessary to employ half the programming profession at any point in time devising conversions from format A to format B.
>
> Michael Kay
> Saxonica

• References:
  • Converting a variety of data formats, containing various kinds ofdata into a common intermediate form
    • From: "Costello, Roger L." <costello@mitre.org>
  • Fwd: Converting a variety of data formats, containingvarious kinds of data into a common intermediate form
    • From: Peter Hunsberger <peter.hunsberger@gmail.com>
  • Re: Converting a variety of data formats, containing various kinds of data into a common intermediate form
    • From: Michael Kay <mike@saxonica.com>