At 10:26 AM 5/5/97 -0700, Gavin Scott wrote:
>Bill writes:
>> I'm not sure I can think of good reasons to do this but I can think of
>> some reasons not to do this:
>>
>> 1.  Performance
>> 2.  Performance
>[snip]
>
>Obviously a man who has never used Query. :-)
>
>I fail to see that this method of access to Image would have any
>fundamental difference in performance than any other client/server
>communication link.  It's just a trick to allow access to the
>Image API through a file-based API.  The performance is entirely
>a function of how you implement it, not the fundamental concept.
>
>G.
>
>
Actually, my past use of Query leads me to the same conclusion!
Any API to Image which leads to serial reads of large datasets
is concerning, from a performance perspective.  If Query can already
do it, and you have a problem with people using it leading to
performance problems, making it easier to do isn't necessarily
a good idea.  This may seem like a data-oriented "security
through obscurity" but, I still say, "why?".

I agree with Gavin (who am I to disagree :-) ) when he says that
it is a function of implementation, just as any API would be.  I
find myself recommending *against* ODBC in many cases, due to the
unacceptable impact this type of activity often has on a system.

Sorry, Birket :-)

Bill Lancaster