Re: [BioRDF] global uniqueness requirement of LSIDs and RDF

Hi Michael and All,
  This is very interesting, especially since similar discussions took place 
in the late 80's about relational versus object oriented.
Same challenge different group.  Thanks Michael for triggering the 
relational world in my mind!!

 How to represent uniqueness of things that span over different contexts.
For example, a street called Mass. Ave. is a single street that goes through 
four cities and maintains the same name.
In fact we have this here in the Boston, MA-USA area.  In relational world, 
as Michael mentioned a new key must be created since the street numbering 
changes in every city and for other variables.
The object world asserted that they can create a single object for the 
street and maybe a method for each city.  However, both approaches agreed 
that a global system uniqueness must be maintained.  Hence, we are still in 
the same discussion.

We all live on the same planet, but we differ in many things, time zone, 
languages, countries, etc.  A representation whether object oriented, 
relational, or LSID, are attempts to represent real world in a digital one. 
In the relational world or object oriented, the digital world was the system 
designed which remained "static" until the next design cycle.  That is why 
it worked.  LSID, is trying to globalize a "dynamic" far reach digital 
world -- I hope I understood that one correctly.  Without putting a context 
and boundaries such representations fall apart very quickly and many start 
shooting holes in them.  We as humans are very good at switching context in 
a dynamic fashion while the digital one doesn't!!

Global uniqueness is hard if not impossible to maintain in a dynamically 
changing and connected world.  What one asserts as correct in one world at a 
point in time might not hold in someone else's world.  Though both could 
learn from their findings once connected and can share experiences.
Hence the goal is not to globalize an LSID, but how to connect LSID's that 
have partial similarities in different contexts over time.  At least that is 
how we overcame such challenges in DiBase, a context-based information 
system.

Cheers
Wafik




----- Original Message ----- 
From: "Miller, Michael D (Rosetta)" <Michael_Miller@Rosettabio.com>
To: "John Barkley" <jbarkley@nist.gov>; <public-semweb-lifesci@w3.org>
Sent: Saturday, August 12, 2006 5:38 PM
Subject: RE: [BioRDF] global uniqueness requirement of LSIDs and RDF



Hi John,

Another version of this problem has existed in the relational world when
importing records from the outside world, which is 'when should an
existing record be updated and when should a new record be created.'
Because the record is coming from the outside world, an alternative key
must be used to see if it already exists, that is, some subset of the
record's properties.

How I've come to think about this is that some properties are intrinsic
to the type of record, for a person, perhaps their SSN if American, and
some are not, such as a person's age.  But even this becomes context
dependent if one wishes to track the state of the person once a year.

The fact that you only have one property for each of your objects
probably oversimplifies the problem.

cheers,
Michael

> -----Original Message-----
> From: public-semweb-lifesci-request@w3.org
> [mailto:public-semweb-lifesci-request@w3.org] On Behalf Of
> John Barkley
> Sent: Tuesday, August 01, 2006 5:31 AM
> To: public-semweb-lifesci@w3.org
> Subject: [BioRDF] global uniqueness requirement of LSIDs and RDF
>
>
>
> The global uniqueness requirement of LSIDs is clear if one is
> talking about
> something like an image. As we discussed on yesterday's
> telecon, if one bit
> in an image is changed, then the LSID must change. My
> question is how does
> this work with an LSID that dereferences to RDF. Consider the
> following
> simple examples:
>
> 1. Suppose in my namespace I create lsid-A in an RDF file
> where lsid-A has a
> single datatype property. Following the global uniqueness
> requirement, if I
> change the value in that datatype property, then I have to
> create a new
> LSID.
>
> 2. Now suppose I create another LSID, lsid-B, that has a single object
> property whose object is the lsid-A from (1). Once again, if
> I change the
> value of lsid-A's datatype property, then I have to create a
> new LSID for
> lsid-A, and also, depending on the meaning I want for lsid-B,
> a new LSID for
> lsid-B with object property the new lsid-A.
>
> 3. Now suppose that I create lsid-C with a single object
> property whose
> object is url-A, also in my namespace. url-A has one datatype
> property. What
> should happen if I change the value of url-A's datatype
> property?  Do I need
> to create a new LSID for lsid-C? I would think the answer
> would be yes.
>
> 4. In (3), url-A is in my namespace. What should happen if
> url-A is not in
> my namespace and the value of its datatype property changes?
>
> Putting these questions more generally:
>
> 1. For RDF, does the global uniqueness requirment mean that only the
> immediate set of properties and their object names/values
> need be unique?
>
> 2. Does it mean that, for RDF, an LSID's closure (of any
> kind) within the
> namepace that I control need be unique?
>
> 3. With RDF, do I have to be concerned about an LSID's
> closure (of any kind)
> in other peoples' namespaces?
>
>
> thanks,
>
> jb
>
>
>
>
>

Received on Sunday, 13 August 2006 17:28:20 UTC