Amazing how my mind holds onto the old, painful, rules. But doesn't remember the rules that make things easier.
On Fri, Jul 11, 2008 at 3:10 PM, Alan Altmark <[EMAIL PROTECTED]> wrote: > On Friday, 07/11/2008 at 02:32 EDT, "McKown, John" > <[EMAIL PROTECTED]> wrote: > > Doesn't a CTC connection require the even/odd address for send/receive? > > No. It requires n and n+1. And even that is just an artifact of the > device driver. There's nothing special about adjacent subchannels on a > CTC. Your IOCP could even configure device number n and n+1 on different > CTC chpids. It just doesn't matter... > > This "even" thing with OSAs and HiperSockets is the vestigal memory of the > original OSA-1s that required the control path to be on an even address. I > think that was lifted in the OSA-2 and was definitely gone by the time > OSA-Express appeared. > > It ranks right up there with PORTNAMEs. They have been optional on z/VM > and Linux for several years, yet I keep finding NEW configurations that > have them coded and people saying "they have to match", which they do, but > ONLY IF YOU CODE THEM! > > Alan Altmark > z/VM Development > IBM Endicott > -- Mark Pace Mainline Information Systems
