>>In that case MSSQL (or any) will never win from native DML on performance, will it ?
>
>I don't see where you're going. The issue is not with SQL Server but with SQL in general. Paging requires that you take a subset of a sequence of rows. In a relational database, the rows of a table are not stored in a specific order. You can't rely on the row being in a physical location. It can be moved depending on the manipulation that is occuring to it.
>
>-Mike

Mike, you are right. But maybe you missed my "(or any").

Further, although I can expect (as I explained myself) that the DB-server won't know anyhthing of the previous call, there might be some reference possible, in fact similar to the VFP native behaviour. Hence, there could be some "recno()" per Index which would do the trick.
Well, anyhow, -and also thinking about the SETS of the former hierarchical DBMSses, I'd think it could have been developed like this.

Nevertheless, your explanation about the all not being stored in a physical order just gives the clue, right ? For that matter, I had already in my mind that the DB-Server had to "know" my ORDER BY from the previous command at least (added with the argument to the SELECT itself). I started with the remark of it all not being possible, and I guess I was putting the question forward against all odds.
But I'm still the newbe, so I was hoping ...

Thanks Mike.