Re: Application logic and Business logic
From: Dmitry A. Kazakov (mailbox_at_dmitry-kazakov.de)
Date: 03/10/05
- Next message: Costin Cozianu: "Re: OO vs. RDB challenge"
- Previous message: Marc Rohloff: "Re: OO vs. RDB challenge"
- In reply to: Alfredo Novoa: "Re: Application logic and Business logic"
- Next in thread: Alfredo Novoa: "Re: Application logic and Business logic"
- Reply: Alfredo Novoa: "Re: Application logic and Business logic"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Thu, 10 Mar 2005 16:12:53 +0100
On Thu, 10 Mar 2005 14:36:23 +0100, Alfredo Novoa wrote:
> On Tue, 8 Mar 2005 16:55:05 +0100, "Dmitry A. Kazakov"
> <mailbox@dmitry-kazakov.de> wrote:
>
>>> See Mikito's post, but it could use the code created by you either.
>>
>>Which shows (as expected) that the relational approach is unable to
>>multiply matrices any close to O(n^2).
>
> You can not derive any conclusion about the performance because you
> don't know anything the implementation.
Well, I forgot about AI connected to RDBMS with a telepathic sensor which
analyses what I want and then writes an optimal program solving it. Come
on, anything is better than searching for elements of a matrix in a 10GB
table.
>>> You have the very same issues if you have a distributed database used
>>> by DBMSless applications.
>>
>>Yes and no. Firstly, many things can be done on the caller's context,
>
> The same as if you use a distributed DBMS with a module that is a DLL.
Which connects to the server via TCP/IP. Come on.
>>because data are not forced to be in DB. Secondly, with advanced ADT I am
>>not bound to one representation.
>
> RDBMS's are infinitely superior doing that.
Start with putting two different types on integers in the same column of a
table!
>> I can have equivalent subtypes: String,
>>UTF8_String, EBCDIC_String, Remote_String. Because the whole system is
>>typed the actual string will be converted only when necessary. The client
>>may work directly in the remote representation.
>
> That's the same if you use a DBMS.
Is there any capable to deal with Unicode? Take any and add new charset
without modifying the DBMS!
>>> For instance container libraries tend to have a low abstraction level,
>>> when you choose a container you are choosing the physical
>>> implementation.
>>
>>Probably because it is known which implementation is better.
>
> But you can not know that because among other things the best
> implementation depends on the data contained in the database at each
> moment.
No it depends solely on the facts of linear algebra. General case matrix
multiplication
>>> If you use an RDBMS you only have a kind of persistent
>>> container and it might be implemented on many ways, and the
>>> implementation might change at any moment without affecting the
>>> applications (except the performance, of course).
>>
>>I see no difference. In fact RDBMS selects an implementation for me at
>>will. I have no influence.
>
> DBA's should be able to select and to change the implementation at any
> moment.
I'm proud of it, but it is not what I want. I want to multiply matrices in
O^2.
>>> Some SQL DBMS's mantain statistics about data and the optimizer uses
>>> the statistics to deduce what is the best access path to the data and
>>> the best algorithms to use at each moment. There are optimizers that
>>> learn how to optimize better with each query.
>>
>>This is because they select a wrong implementation. If I know which
>>implementation is optimal I don't need all that stuff.
>
> There are not optimal data structures. The best structure depends on
> each particular case. That's why physical independence and dynamic
> optimizers are necessary.
That's why DBs are so slow, they optimize. Good to know...
>>OK, what about writing drivers in Prolog?
>
> Prolog is not a very succesfully 4GL, but it would be even more stupid
> to write data management code in Ada.
Actually there are DBMS written in Ada.
>> Unfortunately programming is not
>>about representing facts.
>
> Data management is, and if computerized data management is not
> programming, who cares?
Me, I am a programmer.
-- Regards, Dmitry A. Kazakov http://www.dmitry-kazakov.de
- Next message: Costin Cozianu: "Re: OO vs. RDB challenge"
- Previous message: Marc Rohloff: "Re: OO vs. RDB challenge"
- In reply to: Alfredo Novoa: "Re: Application logic and Business logic"
- Next in thread: Alfredo Novoa: "Re: Application logic and Business logic"
- Reply: Alfredo Novoa: "Re: Application logic and Business logic"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
