Unused selects in SHTOLO

[Günter Staub]

Jeff, Ed, all,

Jeff: I agree to your additions. 

Ed: I don not agree to your following statement:
>I don't see how this request relates to the EXPRESS LRM.  I believe that
>the idea of a "long-form schema" is implicit in the definition of 
>interfacing (clause 11) in 10303-11, not explicit.

The ELRM have an explicit idea of long form: see "Annex G (normative)
Generating a single schema from multiple schemas". This single schema is a
"long form", isn't it?


There were now some discussions around that topic. But I do not see any
decision or any clear position of WG11 on this topic yet.


Kind regards,
Guenter


PDTec GmbH ------------------------------------------
Dr. Guenter Staub             Fon:   +49-721-61844-11
Haid-und-Neu-Str. 7           Fax:   +49-721-61844-19
D-76131 Karlsruhe             Email: staub at pdtec.de
Germany                       Web:   www.pdtec.de


-----Original Message-----
From: jyoung at steptools.com [mailto:jyoung at steptools.com] 
Sent: Tuesday, October 14, 2003 8:26 PM
To: edbark at nist.gov
Cc: rob.bodington at eurostep.com; expressIF at tc184-sc4.org; wg11 at steptools.com
Subject: Re: Unused selects in SHTOLO

>>>>> "E" == Ed Barkmeyer <edbark at nist.gov> writes:

E> Rob is correct that the process of extending EXTENSIBLE SELECT types
produces 
E> artifact select types that have no use.  This may or may not also apply
to 
E> EXTENSIBLE ENUMERATION types.

E> I have no objection to the rule Rob intends, but I think we may want to
be a bit 
E> careful about the wording:

E> (1) The rule should include both SELECT types and ENUMERATION types, and
could 
E> in general apply to any data type that is created by a TYPE declaration
and has 
E> no "role" (is not the data type of any attribute) in the resulting
schema. Any 
E> such data type need not be included in the long form schema, 

There are other possible references to TYPE names besides their use in
the declaration of attribute types.  STRING values compared with the
result of TYPEOF, and the STRING argument to USEDIN, as found in many
local (WHERE) and global (RULE) constraints, for example.  Another
place is LOCAL declarations in FUNCTIONs.  Also, it's probably not
advisable to prune out any TYPE _explicitly_ interfaced, even if it
is not used anywhere in the resulting schema.  (I hope this is not
the very case at issue?)

E> but:

E> (2) When a SELECT data type plays a role, every (interfaced) data type in
the 
E> select list of that select type also plays that role.

E> (3) When a defined data type is included in the long-form schema, its 
E> underlying-type must be included in the schema, even if it plays no role.

Jeff
-- 
Jeff Young (jyoung at steptools.com)
STEP Tools, Inc. / 216 River Street / Troy, New York  12180 USA
Phone: 518.687.2848 x314 (voice) / 518.687.4420 (fax)