Re: the relational model of data objects *and* program objects

mountain man schrieb:
> "Bernard Peek" <bap_at_shrdlu.com> wrote:
>
>

>>It really is a good idea to have some things centralised.

It absolutely clear that a centralized system is easier to create and operate so it is the best choice. However, such a case is only a theory while in practice we need to deel with decentralized system. For example, it would be much more better to have one global post-office for the whole world so that any letter goes to this post-office and then routed to the target address. But it will not work. It would be also very attractive to have one global router which connects all computers. Why not do so if the centralized approach is better? Because it will not work. So decentralization is our payment for having working systems, which are however more difficult to operate and less reliable.

But the issue is actually more complex. In reality there is always some central point in any system and the question is if we use it directly for creating our functionality or we use it it describe our own internal layer (an internal structure) which is then used to describe functions and data. For example, peer-to-peer systems are known to have no central office but strictly speaking it is not so. In this case we have the central part minimized. DNS is also known to be a decentralized system but in this case the center also exists and plays a concrete role. So the central part always exists but may play different role (from minimum to maximum).

alex
http://conceptoriented.com Received on Tue Apr 19 2005 - 15:29:51 CEST