[INFOD-WG] Minutes - INFOD conference call, Thursday, June 28 2007
Fisher, SM (Steve)
S.M.Fisher at rl.ac.uk
Sat Jun 30 19:28:05 CDT 2007
Dieter,
I still don't think it is a problem. The integrity rules are there avoid
references to dead things. When C2 is created P will not be found. If C1
is evaluated after t3 then if it had an explicit reference to P then P
will be found otherwise not.
I will take a read through the use case document to try to make this
more concrete
Steve
> -----Original Message-----
> From: Dieter Gawlick [mailto:dieter.gawlick at oracle.com]
> Sent: 29 June 2007 20:27
> To: Fisher, SM (Steve)
> Cc: INFOD
> Subject: Re: [INFOD-WG] Minutes - INFOD conference call,
> Thursday, June 28 2007
>
> Steve,
>
> The flag is not enough to do the required check.
>
> Let us go through the following scenario:
>
>
> * The is a property instance P created at t1
> * There is a constraint C1 created at t2, P is in the
> solution set of C1
> * P is dropped with 'DISABLE NEW' at t3
> * There is another constraint C2 created a t4, P is in
> the solutions set of C2 - if the DISABLE NEW flag is discarded
> * t1 < t2 < t3 < t4
>
>
> The DISABLE NEW flag is obviously not enough to determine,
> since P remains in the solution set of C1 but has to be
> excluded from the solution set of C2. So, one has to keep
> more information. There are certainly several obvious
> possibility. Adding VALID-FROM and VALID-TO seems to be the
> most sensible.
>
> I am not sure whether this is strictly an implementation
> issue (just a recommendation) or whether we have to mandate
> it. I think we need to mandate it and add it to the
> specifications. Without additional information a GetMetaData
> request would not be able to answer the following question:
> What are the proper instances (and related entries) in the
> solution set of C1 or C2?
>
> Dieter
>
>
> Fisher, SM (Steve) wrote:
>
> * The 'DISABLE NEW' support will
> remain and is
> considered to be useful. A possible
> implementation is to add
> a VALID-FROM and a VALID-TO field to (all?)
> registry elements
> and have a STILL-VALID and a TIME-OVERLAP check when
> constraints are evaluated
>
>
>
> I don't understand this. It says in the spec "There is
> an optional flag
> which can be set to "DISABLE NEW REFERENCES" which
> results in the entry
> being dropped when the last reference to the entry has
> been removed"
>
> The entry is fully visible to anybody currently using
> it and remains
> valid forever and fully functional as long as somebody
> has a reference
> to it.
>
> Steve
>
>
>
> --
>
>
> Oracle Email Signature Logo
> Dieter Gawlick | Architect | 650.506.8706
> Oracle Server Technologies
> 500 Oracle Parkway | Redwood Shores, CA 94065
>
More information about the infod-wg
mailing list