In message <20000703202749.Y360@dr-evil.shagadelic.org>Jason R Thorpe writes
>On Mon, Jul 03, 2000 at 07:53:05PM -0700, Jonathan Stone wrote:

>Right, the point is "for your work". 

The current FD_SETSIZE limit may be satisfactory for *your*
workload. That doesn't mean its satisfactory in general.

The Linux value for FD_SETSIZE (1024) suggest that 256
is not entirely adqeuate for other workloads, as well.

>However, the claim that it
>is insufficient for modern systems (as you, in fact, claimed) 
>is false.

I don't think so.  Even the two datapoints that it's not
adequate for two workloads I know about disproves that.


If its really such an onerous overhead (which I don't buy):
what d'you think of buliding, say, named with a larger FD_SETSIZE?
So that "files 1024" or "files 2048" works as expected?