[plug] Recommending a SQL server

[John Summerfield]

> 
> MySQL seems quite proud of being compliant to various SQL specs, yet you
> rejected it out of hand. Is this a simple case of them majoring in the
> minors, or is there more to it than that?


There is a difference between SQL and "relational." SQL is a programming 
language (a small one to be sure). "relational" is a term used to describe 
the engine behind it.

However the vendor (please interpret that word generously), if it doesn't 
meet the criteria laid down by Codd & Date, it's not relational.

Some years ago (early 80s) I was an ADABAS guru. At the time Codd & Date 
were developing their relational theory, Peter Schnell in Germany was 
writing code that fairly closely matched the developing relational theory.

About that time, Cincom, I think it was, announced their RDBM, a product 
called Supra. At a presntation I attended they asserted (with justice as 
far as I could tell) that no other DBMS (including Db2) was relational.

About that time "relational" became "flavour of the month" and everyone 
(even Ashton-Tate) claimed that their DBMS was relational.


Also at that time DB2 ran like a lame dog; IBM was busily selling IMS and 
inviting people to have a look at DB2 for their less-critical applications.

In Perth (I learned later) SEC wanted to have a look at DB2. It discovered 
it had to upgrade its operating system (from MVS to MVS/XA) and (I think) 
hardware just to run it.

In Canberra, everyone from the ANU to the ABS bought ADABAS; we did not 
present it as relational (it wasn't) but it was somewhat close, and a 
supporting product, Natural, was pretty good.


> 
> I notice that the MySQL test website fails a lot of functions under
> PostGreSQL, and since it marks as "not present" a lot of functionality
> which _is_ present in PostGreSQL I am wondering about its integrity -
> but it might have marked them "in absentia" for a variety of reasons.
> Have you looked at that at all?

As I said, I looked at the specs and didn't like what I saw. I also didn't 
like the fact that its authors think it acceptable to implement decimal as 
real (or maybe double; neither is any good).


> 
> Since you don't rate with PGSQL or MySQL very highly, are there any SQL
> engines that you _do_ like?

I've used mainframe DB2; that's fine. I know the OS/2 version is 
relatively light-weight and I imagine there is no significant difference 
between the OS/2 and linux versions.

I've had a bit of a look at Oracle; setting it up on Linux is a bitch. 
Once it's set up I'd be disappointed if id didn't do a good job - some of 
the world's largest RDBMSs are managed with Oracle.

I had a quick look at Informix; setting that up is relatively difficult 
too. Again, as it's a high-end commercial product, I would expect it do 
perform well.

I've not looked at Sybase at all, but ditto-;)


Understand that with commercial products, you're generally getting a lot 
more than just a file server that implements SQL (which is basically all 
the PG and MySQL are).

When I used DB2 on a mainframe, programs were written in PL/1. Programs 
went through an SQL preprocessor that inspected the source code for SQL 
statements, validated them against the target DBMS and replaced the SQL 
with equivalent PL/1 code. A few lines of SQL could evaluate to pages of 
PL/1 code, and you could be sure the PL/1 code would work.

-- 
Cheers
John Summerfield
http://os2.ami.com.au/os2/ for OS/2 support.
Configuration, networking, combined IBM ftpsites index.