Re: Object databases beat joins (was: Re: ODMG Website?)

"Carl Rosenberger" <carl_at_db4o.com> wrote in message news:<ba5sji$vih$00$1_at_news.t-online.com>...

> > > Walk one pointer!
> > > ...no matter if you have 10 atoms, 2 billion, or 42 fantastillions.
> > > You can get away with one singel read operation.

Like with a RDBMS.

> Direct pointer access performance is constant, flat, always the same:
> FAST!
> ...no matter how much data you store.

That is why it is used in some SQL-DBMSs

> For those of you (Adrian) arguing that a relational database can
> also implement foreign keys as pointers:
> In this case you are violating the no.1 relational principle that
> values (the logical representation) should be independant of the
> physical representation.

It does not violate anything. Data independence is what permits you to implement foreign key constraints as you want.

> The advantage of using direct pointers in object databases, that
> works best for huge databases also works best for limited
> ressources and limited storage capacities:
> PDAs, mobile phones, embedded devices, JavaCard

There are better approaches suitable for RDBMSs, although not yet implemented in industrial strength DBMSs.

> An object database that uses direct pointers to allow navigating
> through objects will achieve the best performance with the least
> ressource consumption. Indices will not be necessary.

Indexes are not necessary in RDBMSs, in fact we can achieve better performance with less resource consumption without using indexes.

Regards
  Alfredo Received on Sun May 18 2003 - 00:27:51 CEST