Re: Dreaming About Redesigning SQL

Some other things I forgot to mention in my last post:

1) Its unfortunate that this thread shows up as two discussions on
google groups.  This is because subscribers to the postgresql hackers
mailing list cannot reply to the other lists.  Its a pity, because
there are some interesting people on that list.

2) I still think there are some problems with the principle of "let
the user only have to know the database at the logical level; separate
the model from the implementation".  It is nice if the user does not
have to know anything about implementation to in order to execute
queries and manipulate data.  Its nice to let the DBA and DBMS be the
one to worry about these things.  But in order to do this properly the
DBA _needs_ to know how the user is _likely_ to make use of the data. 
The user should be able to supply _hints_ to the DBMS to help it to
optimize the way it lays out data, etc.  That's why I used the C++
inline keyword used for small functions.  It's a _hint_ to the
compiler, and it does not affect the results of the program, but it
sure as hell can help if the compiler is not smart enough or lacks
data necessary to make correct decisions.  Rather than insist that the
all query optimization be done by the DBMS, it seems to me that the
thing to do is to make it _unnecessary_, not impossible, for the user
to make things run faster.

3) I guess that's what all the commercial RDMSs are already doing ;)

Another two cents with regards,
Seun Osewa