Re: CYMD dates and SQL -- MIDRANGE-L

I guess it depends on what he means by lowest common denominator. And I
reviewed your comments and think that I could have misunderstood those, as
well.

Al said SQL is a LCD. You talked about DB2. I'm not sure what scope you all
are using.

DDS, the native interface to the database, is clearly not getting much if
any further development. IBM, IIRC, has clearly stated that all future work
is focused on SQL. But if, by DB2 you mean native IO, then I agree with
you. But even in that arena there are things that could help various
performance areas. E.g., AIX, IIRC, has index entries interlaced with the
actual records. This is a structure that could decrease IO, but it would
mean a huge sea change in iSeries. The differences are known and discussed.
Who knows where such discussions may lead?

But in SQL there is much work going on. The new engine (Solomon project) in
v5r2 is an extensive rewrite of everything from the optimizer to access
plan generation (very object-oriented, easily inserted tree nodes, although
not by users), to better statistics on distribution of data in columns,
etc. Much of the work is in getting in sync with the current need to
support dynamic SQL better - as is used so much on the web. IBM chose
embedded, static SQL 20+ years ago, and they were wrong - at least in
today's world. Admittedly, this engine is applied to a subset of queries at
this time - read-only, complex queries - the kind of thing you see in the
TCP-H benchmark, e.g.

Better support for star schema was on the table last year, as I recall.
Certainly SAP needs that support. I have no idea where that is at now.

Attempts are being made to cross-pollinate within IBM. Statistics
collection algorithms from the 390 version of DB2 were adapted, e.g.

Another benchmark that had some influence on things was TCP-W, which can
use ORDER BY and FETCH FIRST ROWS in the subselect of a SELECT. These have
appeared in V5R2.

There's a lot more that could be done. The XPF level of SQL is still the
old stuff - some stuff fairly cobbled together - and rewriting that would
need a lot of justification. Extending the work of Solomon to other types
of queries would be lovely - don't know if it's being done.

So it seems there is (has been?) a lot of activity in the SQL area. DDS is
admittedly stalled out.

Regards

Vern


At 08:14 PM 8/2/02 -0500, you wrote:
>From: <vhamberg@attbi.com>
> > Leif, that is just not true, to the absolute level I
> > infer from your statement. V5R2 has added elements to
> > the SELECT syntax,
>
> > I have to assume you mean something else.
> > > Maybe what Al means is "beyond what they already have made"....
> > > I.e., no changes at all to DB2. DB2 is frozen stiff and unchangeable
> > > (also known as "stable").
>
>I was just trying to interpret Al's statement that IBM is not
>making any more extensions to DB2. If that were true,
>then my statement follows, so you are saying that Al was mistaken?