On Thursday, 04/02/2009 at 12:09 EDT, Erik N Johnson <[email protected]> wrote: > This has especial benefits visa-vi code re-usability. In fact, this > is exactly the design philosophy behind the everything's-a-file model > in UNIX-type operating systems. Of course, that metaphor has been > broken since the advent of the socket, but since then several methods > of extending the file metaphor to representing a network connection > have been conceived.
Diag 0x2A8 was created so that anyone who wants to do QDIO-style I/O can do it without having to infer the operation of the native machine facilities. We can change the implementation without worry about the guest interface, and the device driver author doesn't have to worry about release-to-release compatibility. But since VM TCP/IP already uses the native machine facilities for data transfer, Diag 0x2A8 doesn't help. Not even in the face of guest mobility or the advent of new devices. If a new I/O architecture is required, then it must be because QDIO doesn't cut it. The odds of a such a new I/O architecture fitting the semantics of Diag 0x2A8 are slim. (If it did, then they would probably just use QDIO.) It's the same reason that Diag 0xA4/A8 can't be used with QDIO under the covers - the API semantics are all wrong. Alan Altmark z/VM Development IBM Endicott ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
