Re: Possible bridges between OO programming proponents and relational model
Date: 15 Jun 2006 03:43:01 -0700
Message-ID: <1150368181.077503.232790_at_g10g2000cwb.googlegroups.com>
I know BB won't probably read this thread as he probably used his
magical *twit filter* to discard my post but I want to point out out
the issue of his misunderstanding and bad faith on this thread.
The object of the argument was that I considered physical
representation of SQL table on all current implementation as purely
bidimensional or tridimensional depending on the adressing scheme used
at physical level of RAM implementation. Mister BB has advocated
the opposite stating that physical implementation are multidimensional
and that stated that RAM adresses are linear.
I state that BB is wrong in saying RAM SQL tables representations are
multidimensional...
A multidimensional physical representation of a N dimension relvar
supposes it can be represented through an N-ary coordinate system
equivalent at physical level. Otherly said, any SQL datum location on that physical memory should have then an N-ary system of
coordinates to locate it in RAM.
Such system does not exist.
Most RAM architectures memory is addressed with a 32-bit address (or 36-bit if PAE is enabled). But those addresses are translated by
the memory controller to a physical address that identifies which bank, row, and column the memory actually resides in. In latest Dual
Core Opteron CPU's the memory controller has been directly integrated into the CPU which allows direct physical referencing without need
of translation and took away the need for using linear adressing.
That proves 2 things:
> When stating that RAM is linear, BB has proved that he ignores what a physical memory architecture is. He proved he does not
make a difference between a logical linear adressing scheme at CPU level and a physical adressing scheme at RAM (memory controller). I
clearly pointed out this confusion to him and he chose to persist.
> To be considered multidimensional at physical level a SQL Table should have all the datum embodying it located through an N-ary
physical adressing scheme as opposed to a trinary system. For instance, considering the following example
SQL Table
Datum1, Datum2, Datum3, Datum4 Datum5, Datum6, Datum7, Datum8 Datum9, Datum10, Datum11, Datum12
In a multidimensional physical RAM, it should be possible to represent and locate it at least through N physical coordinates.
--> MEMORY CONTROLER
locates Datum1 at (A1, A2, A3, A4...A-N) locates Datum2 at (A1, A6, A7, A4...A-N) locates Datum3 at (A1, A2, A3, A4...A-N)and so forth....
Needless to such a physical RAM system does not exist. Therefore, it proves SQL Table is NOT multidimensional at physical level. Its
representation and handling is necessarily determined by the RAM adressing scheme used by the memory controler.
To make sure I was not missing anything which could have proved me wrong I have re-read the thread and even checked BB link to try to
understand his point of view. It was a pure waste of time as any of the reading done has been relevant to subject treated. Considering the following:
I consider that unless he acknowledges he was wrong or proves me wrong responding to this question, he just has lost all credibility I was willing to give him. Listenning. Probable answers:
> No response...(let's see if the *plonk* system works ;))
> Irrelevant and off topic responses (talking about something else to divert attention on current subject)
> Insults
> ?
Bob Badour wrote:
> Alfredo Novoa wrote:
>
> > Cimode ha escrito:
> >
> >> The question is whether OO in-memory
> >>mechanisms could support such effort. And how?
> >
> > The answer is evident to me: OO has nothing to offer. Both things are
> > completely unrelated.
>
> Hear! Hear!
Received on Thu Jun 15 2006 - 12:43:01 CEST