"Perry E. Metzger" <perry@piermont.com> writes:

> This makes reasonable sense, and was what I was arguing for before
> until I decided the fight wasn't worth it. For better or worse, most
> systems use SOMAXCONN as "reasonable default" and we may have to stick 
> with that behavior.
> 
> BTW, what does the XPG say?

Although XNS5 permits an implementation to have a (implementation-
dependent) maximum socket listen queue length, it does not specify a
SOMAXCONN (or a differently named but providing similiar
functionality) identifier.  If the backlog argument to listen(2)
exceeds this length, it will be silently truncated; that's all she
wrote.