kohlhase
commented
6 years ago I really like the idea, since
- it further aligns MathML3 and OM2, which is a good thing
- it makes cdbase much more practical
- it gives CDGroups something to do (they were essentially vaporware without a purpose before)
- it is lightweight enough to get it into OM2r2 and we can update the CDs easily.
My only concern is whether a single CDGroup is enough. It was for MathML, since there is only really the mathml.cdg in the purview. Say that I want to send a MathML object via SCSCP, then I would have to have a mathml+scscp.cdg.
Therefore I suggest that we extend David's proposal and add a cdgroups attribute which contains a whitespace-separated list of cdgroup URIs.
davidcarlisle
I would like to propose adding one new optional attribute to (just) OMOBJ
cdgroup="http://.....cdg"taking a URL to a cd group file. This would not affect the abstract OM object (or the binary encoding) but would bring the OM XML encoding in to line with the Strict Content MathML encoding and provide a far more usable way of defaulting URI for individual symbols in an object.
MathML3 says:
https://w3c.github.io/mathml/mathml.html#chapter2_interf.toplevel.atts
We used cdgroup in mathml3 after considering and rejecting adding a cdbase attribute as cdbase is more or less unusable unless all symbols are at the same URL (openmath.org/cd in practice)
If for example you introduce your foobar cd with symbols aaa and bbb at a CDBase of example.com and want an FMP that says aaa=bbb
then you want
apply(relation1.eq,foobar.aaa,foobar.bbb)
In an OM1 style URI-free interpretation that was fine: any application processing the object is assumed implicitly to know some group of CDs and so can interpret this.
But the namespaces-style push to give everything an explicit URI leads to problems: either you set the CDBase of the object to openmath.org to use the standard symbols such as equality then have to explicitly give the example.com CDbase on all the new symbols, or the other way round make the cdbase for the object example.com and mark all the standard symbols with cdbase openmath.org.
The intention with the mathml3 cdgroup attribute is at example.org where you are introducing the cds make a cdgroup that lists (by cdurl) any standard and non standard cds that you wish to use then all your objects just need to have cdgroup=example.org/mycds.cdg and all cd names in the object can be unambiguously tied to any number of different base URI. That is, it rather closely matches the implicit intuition in OM1 that an application "knows" a group of CDs.
This has very low compatibility impact I think and brings the two xml encodings in to line so I'm suggesting milestone 2.0r2 even though it's probably on the limit of syntactic changes that could be considered at a revision.