Re: Possible bridges between OO programming proponents and relational model

Bob Badour wrote:,

//> Cimode wrote:
>
> > I will be more succint in future comments as wordyness seems to be
> > asource of confusion.
> >
> > Bob Badour wrote:
> >
> >>Cimode wrote:
> >>
> >>>To Bob Badour
> >>>
> >>>As a proof of good faith that I am not trying to ellide your questions.
> >>> I will change the quoting. Hope this will clarify...
> >>>
> >>>Bob Badour wrote:
> >>><<What are the two dimensions? Can you name them?>>
> >>>For bidimensional RAM's, the 2 dimensions are RowAddress and
> >>>ColumnAddress
> >>>For tridimensional RAM's (64bit architectures) the 3 dimensions are
> >>>Block Adress, RowAddress, ColumnAddress
> >>>
> >>>Does that make sense? It was written above.
> >>
> >>It doesn't make sense yet. First, you will have to define: "RowAddress",
> >>"ColumnAddress" and "BlockAddress" as part of some coherent
> >>computational model.
> >
> > It does not makes *yet*?. Mister, what is nonsense does not make sense
> > *ever*.
>
> I don't recall saying that the above is nonsense. I recall saying it
> cannot make sense until you define the terms "RowAddress",
> "ColumnAddress" and "BlockAddress". Until that time, it is merely
> meaningless.//
Well in substance, you are saying it is nonsense until it would be defined. This is totally incoherent.
RowAddress, ColumnAddress and BlockAddress are explicitely defined through terminology and do not need a definition.

>
> > Besides when was the last time you checked memory architectures?
>
//I suggest you consider the name of this newsgroup. It is not  comp.intel.architecture.64bit and it is not comp.intel.mmu.64bt// You have totally ellided that question.
The name of this newsgroup is comp.databases.theory and it is a ground for discussing all theory related to databases including implementation theory.

>
> If you want to discuss a specific computational model, you will have to
> provide some sensible definition of that model. Even with segmented
> memory, the address space is linear. Even with virtual memory, the
> address space is linear. Even with paged memory, the address space is
> linear.
You keep bringing computational model again. You see models everywhere. While I told you that segmented memory has nothing to do with physical addressing scheme for relative position of data.
>
> Are you suggesting that current memory architectures are non-segmented,
> non-virtual, non-paged, non-linear architectures? Are you suggesting
> that this would somehow become relevant to the theoretical computational
> model?
I suggested that this high level not low level.

>
> While you initially seemed to show some promise, you are beginning to
> convince me you are nothing but a crank after all.

I do not need your paternalistic to educate myself. I know how to read and that's amply sufficient. And given theamount of incoherence you demonstrated so far, I will take that as a compliment.

> Current 64 bit memory architectures already support this
> > adressing scheme without the need to inventing a computational model
> > (For whatever it maybe). What do you call *linear*
>
> I call a 1-manifold "linear". What do you call linear? What do you call
> "bidimensional" ? (Other than some imagined implementation of an SQL
> table, that is.)
>
> In a computational model, I call any addressing scheme requiring single
> pointers "linear". On a CPU, I call any address bus consisting of a
> fixed number of external pins representing the state of some CPU
> register "linear".

This is what I assumed. Your are confusing an absolute addressing scheme a relative adressing scheme. You assume that all adresses in memory are directly refered through their absolute memory only. Why do you bring CPU into this?

> >>Obviously, you are trying to define some computational model for the
> >>physical implementation of a dbms. You apparently want to discard the
> >>linear memory addressing scheme used on almost all computer
> >>architectures for a different addressing scheme.
> >>
> >>Once you completely define the computational model, what you say might
> >>make sense. Until then, you are just using meaningless terms.
> >>
> >>Apparently, in your addressing scheme, one may not access memory through
> >>a single pointer. Instead, one must use three pointers.
> >
> > This comment is the most Interesting you have posted so far. What is
> > linear to you?
>
> See above. What is linear to you? What is bidimensional? What
> computational model do you want to discuss? What property of that
> computational model do you think an OO computational model might improve?
If I am not mistaken what you call "Linear" adressing scheme is what I call "absolute" adressing scheme. I have not refered to a computational model, you did. My hope is that OO *some* logical and physical mechanisms such as a more coherent use of pointers would allow totally dissociate physical layer from logical layer. IN the same spirit, inheritance being distinctive of OO may allow more efficient and coherent use subtypes/supertypes. As some posters evoked, tons of works by serious audiences have been addressing that issue (Oh yeah you are going to start saying that DATE is a crank). Received on Mon Jun 05 2006 - 10:59:41 CEST