- From: Christian de Sainte Marie <csma@ilog.fr>
- Date: Tue, 29 Jan 2008 11:02:20 +0100
- To: Michael Kifer <kifer@cs.sunysb.edu>
- CC: Igor Mozetic <igor.mozetic@ijs.si>, public-rif-wg@w3.org
Michael Kifer wrote:
>
> SystemA RIF SystemB
> p(foo->2,bar->"abc") ----> p(foo(2),bar("abc")) -----> p(foo(2),bar("abc"))
During a private conversation, someone suggetsed that conformant
translators from systems using named arguments Uniterm to RIF could be
required to order the slots (e.g. lexicographically, but any convention
would do).
That way, named argument Uniterms would be passed around as standard
Uniterms, without ambiguity and without impacting systems that do not
use named arguments uniterms. And the (ordered) list of the slot names
(or TERMs representing them) could be passed as meta data (or whatever
we call information that can be ignored but that must be preserved and
passed along), so that consumer systems that use named argument uniterms
would be able to reconstruct the slotted form (and others would just
ignore it):
SystemA RIF SystemB
p(foo->2,bar->"abc") ---> p[bar,foo]("abc", 2) ---> p(bar->"abc",foo->2)
|
| SystemC
-----> p("abc",2)
The burden would be on the implemenors of tranlators from systems with
slotted uniterms to RIF, of course. But (i) the burden is "only" to
order the slots, and (ii) they seem to be a minority (that is, the
majority of implementors would not be impacted).
Would that work?
Christian
Received on Tuesday, 29 January 2008 10:02:21 UTC