Posted by --CELKO-- on 02/06/06 16:25

>> imagine if you will 5 tables. 3 of them are over a billion rows .. <<

Since you did not post DDL, I do not see how they are related or what
the query is. When you say "children" I assume you mean referencing
tables? But in the real world, if I am dealing with an RDBMS that has
billion row tables, I would not use SQL Server at all.

This is a job for Teradata, SAND, DB2, etc. I would want to have at
least table partitioning and parallelism.

>> sql embedded in code is really bad. <<

Unh? I never said anything about embedded SQL. I fact, I explicitly
said that VIEWs and Stored Procedures serve different purposes.

We used embedded SQL a lot in the early days wtih Cobol, PL/I and C (in
fact, I have an example in C in my book INSTANT SQL). The precompilers
were easier to use and more portable than the assorted APIs. You could
get an executable that did not reveal source code the guys that bought
your package. But the source code showed you both the host code and
the SQL code when you maintained it.

Then along came ODBC, ADAPI and SQL/CLI which match a teired
architecture. embedded SQL fell out of favor. I doubt most current
programmers would recognize this:

EXEC SQL INCLUDE SQLCA;
EXEC SQL BEGIN DECLARE SECTION;
host_name character_string(20);
host_emp_number integer;
EXEC SQL END DECLARE SECTION;
EXEC SQL WHENEVER SQLERROR STOP;
EXEC SQL CONNECT frans;

There is a website at
http://www.csc.liv.ac.uk/~frans/Notes/embeddedSQL.html#syntax if you
are interested.

[Back to original message]