Re: [ubl-cmsc] The "Phase 1 context methodology" proposal

[Eduardo Gutentag <eduardo.gutentag@sun.com>]

Eve L. Maler wrote:

> I took an action to explain the proposal that I came up with for a 
> Phase 1 CM deliverable. I'm also supposed to try to set up a CM 
> telecon to keep the discussion going. (There were other big CM 
> developments last week as well...) Matt, s'okay with you? If so, here 
> are the times I'm free in the coming few days; please respond with 
> your own subset of these times and I would be happy to set up a phone 
> call:
>
> Fri 14 Jun: 7am-10am PT / 10am-1pm ET / 3-6pm UK / 4-7pm Eur 

 From 8amPT on

>
> Mon 17 Jun: 7am-10am PT / 10am-1pm ET / 3-6pm UK / 4-7pm Eur 

Not available

>
> Tue 18 Jun: 7am-11am PT / 10am-2pm ET / 3-7pm UK / 4-8pm Eur 

Not available

>
> Thu 20 Jun: 9am-11am PT / 12n-2pm ET / 5-7pm UK / 6-8pm Eur 

 From 10amPT on

>
> Fri 14 Jun: 7am-10am PT / 10am-1pm ET / 3-6pm UK / 4-7pm Eur 

 From 8amPT on

>
>
> Eventually I will have to write up the proposal properly (and give it 
> a cutesie name; nothing has stuck yet :-), but I wanted to try and get 
> the bare bones out there as soon as I could. Here it is:
>
> Summary:
>
> . For Phase 1, do not yet enable the description of *deltas* between
> less-contextualized and more-contextualized types and schemas (context
> rules). Instead, enable merely the *description of the actual
> contexts* that apply to hand-derived schema modules built on top of
> UBL. Allow these descriptions to be "defaulted" somehow for all types
> in a schema module, for ease/economy of context value assignment.
>
> Consequences:
>
> . Makes contextualized schemas stay strictly compliant with XSD
> derivation. (And likely pushes us in this direction for future CM
> phases...) Allows us not to have to re-solve all the problems that
> XSD currently solves, such as worrying about how to set the namespace
> of the constructs in the resulting contextualized schema.
>
> . Pushes the work of generating derivations onto fairly knowledgeable
> schemographers. (Since there are tools and tutorials extant for XSD
> work but not yet for context rules design, this doesn't seem like much
> of a cost.)
>
> . Will probably encourage the creation of contextualized schemas
> overall, possibly resulting in a more confused picture of what modules
> get registered and what their relationships are to each other.
> However, getting this input will be a real win for our work on Phase
> 2.
>
> . Will result in the context descriptors for contextualized schemas and
> for the base UBL Library modules taking exactly the same form, which
> is nice.
>
> Work to be done:
>
> . We need to fully solve the general problem of documenting applicable
> contexts in order to allow hand-coders to add their context
> descriptions to their own types. We discussed this at some length
> in the NDR meeting last week and made some progress.
>
> . We need to figure out the defaulting rules of context descriptors that
> appear "globally" in the prolog of a schema module. It must be
> possible at any time to "compute", given a stack of schema modules,
> the applicable context for any object class.
>
> . We need to fully flesh out the rules for allowable contextualizations,
> much as XSD defines the rules for allowable derivations. The context
> descriptors will be metadata on types that forms a phantom context
> hierarchy; what are the rules for context derivation? Eduardo has
> already proposed one about properly specializing the context value
> hierarchy (see the NDR F2F minutes). We will probably also need
> rules about computing from "additive" contexts (e.g.,
> geo="Americas+Asia") and ranges of contexts.
>
> I would say that, if this proposal is acceptable to the fully 
> constituted CM group, we need to produce this deliverable *fast* and 
> we would need to promote it in much the same way as the code list 
> recommendation, since both apply mostly to external producers and not 
> just UBL proper.
>
> A random observation: If formal mathematical comparison of XSD modules 
> were easy to do, in the fashion that RELAX NG is amenable to this, 
> then heck -- the context rules themselves could possibly be computed 
> from the comparison of two contextually related schema modules!
>
> BTW, I am taking cutesie name suggestions.
>
> Eve
>

-- 
Eduardo Gutentag               |         e-mail: eduardo.gutentag@Sun.COM
XML Technology Center          |         Phone:  (510) 986-3651 x73651
Sun Microsystems Inc.          |         1800 Harrison St. Oakland, CA 94612