RE: [PATCH v5] introduce vfio-user protocol specification

[Thanos Makatos]

> -----Original Message-----
> From: John Levon <levon@movementarian.org>
> Sent: 07 November 2020 12:26
> To: John G Johnson <john.g.johnson@oracle.com>
> Cc: Thanos Makatos <thanos.makatos@nutanix.com>;
> benjamin.walker@intel.com; Elena Ufimtseva
> <elena.ufimtseva@oracle.com>; tomassetti.andrea@gmail.com;
> jag.raman@oracle.com; james.r.harris@intel.com; Swapnil Ingle
> <swapnil.ingle@nutanix.com>; yuvalkashtan@gmail.com;
> konrad.wilk@oracle.com; Kanth.Ghatraju@oracle.com; qemu-
> devel@nongnu.org; tina.zhang@intel.com; ismael@linux.com;
> alex.williamson@redhat.com; Stefan Hajnoczi <stefanha@redhat.com>;
> Felipe Franciosi <felipe@nutanix.com>; xiuchun.lu@intel.com; Marc-André
> Lureau <marcandre.lureau@redhat.com>; Raphael Norwitz
> <raphael.norwitz@nutanix.com>; changpeng.liu@intel.com;
> dgilbert@redhat.com
> Subject: Re: [PATCH v5] introduce vfio-user protocol specification
> 
> On Thu, Nov 05, 2020 at 05:50:27PM -0800, John G Johnson wrote:
> 
> >     The idea behind the version IDs is to identify incompatible protocol
> > changes as major versions, and compatible changes as minor versions.
> What
> > would be the purpose of the third version type?
> 
> Well, like any patch version, it'd be for identifying versions on the other 
> side
> for reporting, debugging purposes. Not imply anything about the protocol
> version. But it's not a big deal.
> 
> >     The thing that makes parsing the JSON easier is knowing the version
> > beforehand so the parser knows what keys to expect, so I’d like to
> promote
> > major and minor to separate fields in the packet from being embedded at
> an
> > arbitrary points in the JSON string.
> 
> I agree that'd be a sensible change (and then I wonder if the little bit of 
> JSON
> is actually useful any more).

The reason why the JSON string exists is that it simplifies adding information 
to the version, should we ever need to.

> 
> > >> So can we switch it now so the initial setup is a send/recv too?
> > >
> > > I'm fine with that but would first like to hear back from John in case he
> objects.
> >
> >
> >     I think I write that section, and just switched client and server.  The
> code
> > is written as client proposes, server responds; this is the better model.
> 
> Hah, great, thanks.
> 
> regards
> john