Michael VanLoon writes:

>I think I'm getting confused trying to follow this.  How do you define
>"one", here?

As in "the person writing the driver".

>I was under the assumption that the 80186 on my BusLogic
>card did all the waiting and sending for me.

That's something entirely different.  The BusLogic card still can't
send more then one command to the target if it's not using tags.

That could make a difference when you send a command to write block
N and then a command to write block N-1 (or similar cases).  In the
untagged case you've just missed the opportunity to write both blocks
in less then one revolution of the disk.  OTOH you might have been
lucky and prevented a file system inconsitency from reordering writes.

Of course, this depends on how often this happens, whether the drive's
firmware is actually smart and fast enough to take advantage of it,
and whether you want your carefully ordered writes to be reordered by
some disk firmware that's optimised for DOS or similar Lameness.

Given that BSD drivers are supposed to sort disk requests to yield a
good seek pattern, that modern drives have at least track sized read
caches (and older drives don't support command queuing anyway) and
sometimes write caches it doesn't seem like a big win to me.

I guess, I'll do some benchmarks in the near future.
-- 
Christoph Badura	bad@flatlin.ka.sub.org

You don't need to quote my .signature.  Everyone has seen it by now.
Besides, it doesn't add anything to the current thread.