Title: Message
A few
problems I see with the new schemas. I'm looking at "reusable" now...
1.
Many of the globally defined elements are still carrying the misleading
annotation. This was the problem Tim was trying to fix with last
Friday's (long tag names) release. For instance, the (global)
CompanyRegistrationID element carries the documentation from "Party Tax
Scheme.Company Registration.Identifier". The documentation annotation on
the element should be empty. Note that the element that ref's this one is
documented correctly -- see the definition of
'PartyTaxSchemeType'.
2.
It's a bit hard to see this one, and it's more of a fine point -- not a
showstopper but the schema is defining global elements that are never
used (ref'ed). For instance AccountsContact is defined but never
ref'ed. I believe this is because the Perl is (correctly) following rule
R9 from NDR (R9 says you've got to gen one global element for each complex
type). It seems to me now that R9 is superfluous. R13 on its own
necessitates local elements ref'ing global elements. R13 specifies a
_sufficient_ set of element declarations. R9 prescribes extra ones.
Again, look at AccountsContact: R9 says we have to have an element defined for
it (alias)... but it's never used in any content model. Who is to say that
when/if it is used in some content model -- possibly by a document schema like
Order, that the property term will be 'AccountsContact'? I move for doing
away with R9, adding a little text to R13 to make it explicit that we need to
create the ref'ed element declaration and changing the Perl at some point in the
future to reflect the changes.
Now
for some fun statistics on Gunther's latest schemas. I'm looking just
at the "reusable" schema:
count(//xsd:complexType) that is the number of complex types 91. To
this number we should add the 15 CCTs giving a total of
106.
count(//xsd:element[count(@ref) = 0]) that is the number of globally
defined elements: 275
count(//xsd:element[@ref]) that is the number of local elements
ref'ing globally defined ones: 570
So the
average complex type is used 275/106 = 2.6 semantically distinct ways.
That is, on average each complex type (structure) has about 3 different
"meanings". The distribution is skewed though. Each ABIE
had exactly on element defined in terms of it, whereas the typical CCT
might have many (e.g. cct:IdentifierType has about 20 element defined in terms
of it). There's more "aliasing" of the more "primitive"
types.
The
average element is used (ref'ed) at 570/275 = 2.1 places in the
vocabulary. This is the central result of the latest script
revisions. Yesterday that number was 1, today it's 2.1. To my mind
this could be interpreted as: a user of the UBL vocabulary has half as many
names to learn and remember.
-----Original Message-----
From: Stuhec, Gunther
[mailto:]
Sent: Wednesday, January 22, 2003
6:44 AM
To: UBL-LCSC ();
Subject: [ubl-ndrsc] Actual version
of UBL library as well as the perl script
Hello all,
I apologize that I'm a little bit late. But fixed
some another bugs within the perl script. That needed a little bit more time.
Now, I generated with the modified and bug fixed perl script the libraries as
discussed.
--> I deleted "Details" from each aggregate.
--> The namespaces are according our
conventions.
Kind regards,
Gunther
<<baseline_21-01-03.zip>>
