> From: Jason Thorpe <thorpej@nas.nasa.gov>
> Date: 1999-06-28 17:30:41 -0700
> To: Darren Reed <darrenr@reed.wattle.id.au>
> Subject: Re: accept(2) behaviour.
> Cc: tech-net@netbsd.org
> Delivered-to: tech-net@netbsd.org
>
> On Tue, 29 Jun 1999 10:06:01 +1000 (EST)
>  Darren Reed <darrenr@reed.wattle.id.au> wrote:
>
>  > > Why not add a flag?
>  >
>  > Well, that can be done, sure.  But looking at what's there, it  
already seems
>  > like PR_CONNREQUIRED supplies the same information as another  
flag saying
>  > "listen/accept supported".  I guess the question is would using  
that flag
>  > for that purpose be a misuse of it ?
>
> Not only do I think it's a misuse of "connection required", but what 
> about e.g. connected UDP sockets?  Can you listen(2) on them?

The PR_CONNREQUIRED flag is not set for UDP, and doing a 'connect'  
on a UDP socket wouldn't affect this:  listen() is not supported on  
UDP sockets, independent of their connected state (but wouldn't it be  
kinda weird doing a listen on a connected socket anyway?).

What do you see as misuse in this case?  It seems like the 'listen'  
flag would be turn on in precisely those cases where the  
PR_CONNREQUIRED flag is set, no?

Regards,

Justin

--
Justin C. Walker, Curmudgeon-At-Large *
Institute for General Semantics       |
Manager, CoreOS Networking            | When crypto is outlawed,
Apple Computer, Inc.                  | Only outlaws will have crypto.
2 Infinite Loop                       |
Cupertino, CA 95014                   |
*-------------------------------------*-------------------------------*