> So I'm reading, apart from loading a whole row (which could be a
> problem for records with BLOBS, nested datasets etc), there was no
> major performance penalty.

To me, it certainly felt like there was a penalty. As to the reason
for the penalty, I suspect that your suggestion of multiple table
joins was part of it, but I'm not in a position to investigate or
comment in any detail now.

More generally, abstraction has its cost. That cost is usually not
too significant for in-memory operations on modern CPUs. It can even
work with intensive data accesses in non-cached memory or *even* on
fast local disks. But if the cost of abstraction is overlaid on
relatively-slow network database connections (i.e. any normal, live,
real-world database application), then at times it can become too
much to bear.

My personal feeling from a performance/value point of view is that
this whole area -- much like .NET itself -- is where Windows 1.0 and
2.0 were in the late 80s. Win3.1, let alone Win95, is still a ways
off yet.

But Win3.1 did happen, eventually -- mainly because the hardware
caught up with its demands. So when we all have database servers
employing well-exploited parallel processors and super-fast disks and
unsaturated +Gbit LANs, all may fall into place.

But, as I've observed recently, this isn't going to happen quite as
fast as we've become used to.

cheers,
peter


_______________________________________________
Delphi mailing list
[email protected]
http://ns3.123.co.nz/mailman/listinfo/delphi

Reply via email to