RE: [xacml] Proposal of Model for comparing two sequences

[Michiharu Kudoh <KUDO@jp.ibm.com>]

Daniel,

>Only First is bad because our data model does not guarantee order,

I thought that our data model supports sequences as described in Function
draft 0.8. I'm not completely sold on our current function set to have
bunch of functions such as "string-first", "integer-first",
"decimal-first", "date-first", "time-first" etc. For the reiteration of the
discussion, I could not follow the discussion when you and Polar had. But
through the definitions in the current draft and the discussion on the
list, I finally recognized the issue from my viewpoint.

Michiharu Kudo

IBM Tokyo Research Laboratory, Internet Technology
Tel. +81 (46) 215-4642   Fax +81 (46) 273-7428




                                                                                                                                                    
                      Daniel Engovatov                                                                                                              
                      <dengovatov@cross        To:       Michiharu Kudoh/Japan/IBM@IBMJP, "'XACML '" <xacml@lists.oasis-open.org>                   
                      logix.com>               cc:                                                                                                  
                                               Subject:  RE: [xacml] Proposal of Model for comparing two sequences                                  
                      2002/09/12 10:32                                                                                                              
                                                                                                                                                    
                                                                                                                                                    



>Polar

>First, my proposal is meant for applying only to the comparison operator
>(meaning that it returns a boolean value), not to arithmetic computation
>functions such as integer-subtract. I think we had better distinguish
these
>two types if we support sequence in our function model. Personally,
>arithmetic functions are supposed to implicitly have "OnlyFirst"
>ComparisonBase.

Only First is bad because our data model does not guarantee order, thus
evaluation
outcome will not be predictible - we went through this discussion
with *-first functions.

>If you think that comparison base for arithmetic functions are redundant,
>it can be omitted. Then only we have to do is to add a text that the
>comparison semantics for integer-subtarct function is "OnlyFirst". Does it
>make sense?

That was in the first iteration of function specification - implicit
runtime
check
of the return sequence size to be appropriate (indeterminate being returned
for
a wrong cardinality).  I still think this is OK, but had to agree with
Polar
that
compile time type safety is worth it..

We seem to reiterate exactly the same issues in different form over and
over..
Indeed, not very obvious..

daniel;