To add to this:

I sometimes have the impression, that PM's aren't delivered at once, but rather after some time.

...oh, I see. Well, no wonder, I didn't understand. This acronym isn't usual here in Germany 

If your column don't allow for NULLs, you have to change them to allow for NULLs, or insert a dummy value. But what are you really after???

I don't understand your reference to JC?

But I agree Yoda's answer should also be interesting.

INSERT INTO tbl2 (fname) SELECT fname FROM tbl1 WHERE ref#='123'

Btw, the # sign isn't really good to be used as a column name. I have seen several threads here, where...

Google smart and google in the Groups, not the web and you'll find

DBCC OPENTRAN might be a start

The fact that this is default behaviour of SQL Server, does not mean that it cannot be changed. Just explicitely created the PK as nonclustered, like so:

create table #t

(

 c1 int primary...

See, if this helps:

http://www.microsoft.com/resources/documentation/sql/2000/all/reskit/en-us/part3/c1161.mspx

Or search the fora here. This is a FAQ in almost any SQL Server community.

Hm, maybe Joe Celko knows...

And if he answers, I would like to know why ANSI invented NULL.

The result is the same. Actually you should find, when you compare execution plans, that such statement are execute identical. The optimizer should restate it to using JOINs. However using JOINs is...

What does this return?

select databasepropertyex('your db','recovery')

IIRC, this number is 4 or 5. Beyond this the optimiser will choose tablescans, because of the sheer amount of possible permutations.

Even better is to use a language independant format.

Here are some very good thoughts by SQL Server MVP Tibor Karaszi on this topic: http://www.karaszi.com/sqlserver/info_datetime.asp

...oh, and I forgot, someone who calls himself "admin", needs to be more responsible. This should not happen. You might want to give him a more or less friendly slap...