Re: [Qemu-devel] [RFC] docs: vhost-user: add in-band kick/call messages

[Johannes Berg]

On Thu, 2019-09-12 at 14:22 +0200, Stefan Hajnoczi wrote:
> 
> The vhost-user spec is unclear and inconsistent.  Patches are welcome.

:)

> A footnote describing the old terminology would be necessary so that
> existing documentation, code, etc can still be decyphered when the spec
> changes the terminology :).

Yeah.

But we (you) would have to decide first what the terminology actually
*should* be :-)

I'd have said something like "host" and "device", but then "host" can
get confusing in the context of qemu, is it the host that the VM is
running on? It's the VM that's hosting the device, but that's a bit confusing 
in this context.

Client/server might work, but it's not very obvious either (**), and
quite a bit of the text actually gets it wrong right now, so it'd be
very confusing to swap that and have a footnote or similar indicate that
it was described the other way around previously ...

Calling it master/slave as the text does now is a bit confusing to me
because it has DMA (of sorts) and that's called "bus mastering" in most
other contexts, but perhaps that's what would work best nonetheless,
though I'm not really a fan of that terminology.

Perhaps master/device would be nicer, getting the term "device" in there
somewhere would seem appropriate, even if it's not really actually a
device, but from the view inside the "master" VM it is a device...


(**) Btw, is it really true that there's a way to have the device make
the connection to a listening unix domain socket, as implied by the
spec? I cannot find evidence for such a mode in any code...

johannes