Re: Extending my question. Was: The relational model and relational algebra - why did SQL become the industry standard?

Steve,

My idea of a useful logical data model is one that costs less to own, is easier to learn, reduces the risk of human error and requires fewer expert level users to maintain. I am astounded you would have a different idea of utility.

There are three levels of discourse: conceptual, logical and physical.

The conceptual level of discourse involves information understood by humans but not necessarily in a form suitable for machine processing.

The logical level of discourse involves abstract data in a form suitable for machine processing but not limited to the properties of any particular machine.

The physical level of discourse involves an actual machine or machine type.

A multiset model requiring knowledge of physical properties of a particular machine or machine type is not strictly speaking a logical model--it combines both logical and physical into one. The problems introduced by doing this are legion.

For instance, some information is only accessible implicitly from the position of some other data. This means that a dbms based on such a model must manage both data and position--it's really a dbPms. Users must know how to access both data and position. Both the dbms and the users require a more complex language to manipulate and communicate both data and position. The requirement to maintain both data and position means the dbms has fewer options for optimizing data access and for optimizing position access. For example, the requirement to manage position limits the transformations available for rewriting data access queries.

Given the additional complexity, users will require greater expertise to successfully use the DBMS and will always face an increased risk of error.

"Steve Kass" <skass_at_drew.edu> wrote in message news:b2vbkd$r3b$1_at_slb3.atl.mindspring.net...
> We don't have the same idea of what it means for a logical model
> to be useful, it seems. I define the utility of a logical model
> in terms of its relationship with a real-world model (it does a good
> job of logically representing the real-world model), and its relationship
> with a physical implementation (it admits a straightforward faithful
> physical implementation). Since I don't know what your definition
> of utility is, I don't know whether bags would clear the bar.
>
> Steve
>
> Bob Badour wrote:
>
> >All of your multiset examples involve external physical representations.
> >Why, then, are you suggesting a use for multisets as a logical
> >representation?
> >
> >"Steve Kass" <skass_at_drew.edu> wrote in message
> >news:b2ulsp$10v$1_at_slb2.atl.mindspring.net...
> >
> >
> >>Here are a few, including examples where <item, multiplicity> is a
> >>natural conceptual model and where <item, item, item> is.
> >>(also see my post previous to this, answering Bernard)
> >>
> >>An inventory.
> >>When I inquire at a bookstore as to the availability
> >>of a book, someone looks it up in a database and may report "we have
> >>three in stock". If I purchase one, the physical implementation changes
> >>a "3" to a "2". It doesn't delete a fact from a table. Similarly, my
> >>
> >>
> >local
> >
> >
> >>pharmacy probably maintains an inventory of medications on hand. I
> >>doubt that 1,237 tablets of alprazolam are represented by that many
> >>entries in a database.
> >>
> >>Ticket sales.
> >>"We sold 325 adult tickets and 104 child tickets today."
> >>The tickets are separate physical entities, and when a family purchase
> >>several, a machine issues several identical objects (they may well be
> >>numbered, unfortunately), so this example is one where there is some
> >>representation of the multiset as individual but identical items.
> >>
> >>A receipt.
> >>SQL Burger $3.29
> >>Big Int $1.24
> >>Table Service $.99
> >>SQL Burger $3.29
> >>SQL Burger $3.29
> >>
> >>Groceries.
> >>"Could I have two dozen jumbo shrimp, please?"
> >>"and a dozen eggs?"
> >>"and 5 pork and scallion buns?"
> >>
> >>Perhaps I gave you the impression that I had something more
> >>subtle in mind, but these are the kinds of things for which I think
> >>multisets, sometimes conceived as sets of <item,multiplicity> pairs
> >>and sometimes as true bags, are useful.
> >>
> >>SK
> >>
> >>
> >>
> >>>>
> >>>>
> >>>Steve,
> >>>
> >>>Could you provide a practical example that might help me in wrapping my
> >>>
> >>>
> >mind
> >
> >
> >>>around the notion of utility for logical models of items with
> >>>multiplicities?
> >>>
> >>>I'm afraid I'm having problems understanding how one can even have some
> >>>sense of determinancy of what constitutes a multiset in contrast to a
set
> >>>without some implicit logical mapping to identity. In the mind's eye,
> >>>
> >>>
> >the
> >
> >
> >>>very basis for contrasting a multiset from a set, or vice versa, is
> >>>dependent on our very notion of identity.
> >>>
> >>>For example, if I see {1,1,1,1,1}, I would have a tendency to describe
it
> >>>
> >>>
> >as
> >
> >
> >>>a collection of integer 1 values with a cardinality of five. In the
> >>>
> >>>
> >process
> >
> >
> >>>of synthesizing my description, I find that I implicitly assign
> >>>
> >>>
> >cardinality
> >
> >
> >>>to each member even though set theory would reduce this to {1}. Thus,
I
> >>>
> >>>
> >can
> >
> >
> >>>distinguish between the two collections.
> >>>
> >>>What am I missing?
> >>>
> >>>I guess the root of my confusion lies in the fact that I don't see how
we
> >>>relate to anything in the real world without trying to apply some
notion
> >>>
> >>>
> >of
> >
> >
> >>>identity in a logical sense.
> >>>
> >>>
> >>>
> >>>>>
> >>>>>
> >>>
> >>>
> >>>
> >>>
> >
> >
> >
> >
>
Received on Thu Feb 20 2003 - 03:03:36 CET