Re: Collections: Sets, bags or something else?

Asir Vedamuthu wrote:
>
> > that "collection" here means "unordered collection with
>
> >duplicates allowed", informally known as a "bag".
> >Is this the intended meaning?
>
>  
>
> Yes
>
>  
>
> >If the intended meaning is to allow duplicates, is there
>
> >any special meaning to the same /alternative/ appearing more than
>
> >once in a policy
>
>  
>
> No
>
In that case, define policies as /sets/ of alternatives, as there
appears to be no reason to deviate from that well-established abstraction.

In general, it seems risky to use off-the-beaten-path concepts like bags
unless there is some compelling reason to do so.  Intersection of
alternatives is a case in point.  If alternatives were sets, not bags,
the definition of intersection could (AFAICT) say "set union" or "set
intersection"(whichever was appropriate) and there would be no room for
misunderstanding.  As it is, you're effectively defining "bag
intersection" or "bag union", or perhaps something else, on the fly.

In the case of policies as sets of alternatives, it would be clear that
order and multiplicity don't matter, only what is or isn't in the set --
in this case, what alternatives are or are not available.
>
>  
>
> We hope this helps.
>
>  
>
> Regards,
>
>  
>
> Asir S Vedamuthu
>
> Microsoft Corporation
>
>  
>
>  
>
> *From:* public-ws-policy-request@w3.org
> [mailto:public-ws-policy-request@w3.org] *On Behalf Of *David Hull
> *Sent:* Monday, April 30, 2007 10:24 PM
> *To:* public-ws-policy@w3.org
> *Subject:* Collections: Sets, bags or something else?
>
>  
>
> A follow-up to my previous:
>
> The spec appears to carefully use "collection" and not "set".  This,
> together with the absence of expression equivalence rules like /a+a=a/
> and /a*a=a/ and the note that assertions of the same /type/ may occur
> in an alternative, suggest that "collection" here means "unordered
> collection with duplicates allowed", informally known as a "bag".
>
> Is this the intended meaning?  It's not unheard of to use "collection"
> to mean "set" (i.e., duplicates are not considered).  If the intended
> meaning is to allow duplicates, is there any special meaning to the
> same /alternative/ appearing more than once in a policy (as opposed to
> the same /assertion/ (type?) appearing more than once in an
> alternative, which behavior is out of scope).
>