Re: [PATCH 10/27] docs: Libxl migration v2 stream specification

[Ian Campbell <ian.campbell@citrix.com>]

On Wed, 2015-07-08 at 14:49 +0100, Andrew Cooper wrote:
> On 16/06/2015 14:58, Ian Campbell wrote:
> > On Mon, 2015-06-15 at 14:44 +0100, Andrew Cooper wrote:
> >> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
> >> +EMULATOR\_CONTEXT
> >> +----------------
> >> +
> >> +A context blob for a specific emulator associated with the domain.
> >> +
> >> +     0     1     2     3     4     5     6     7 octet
> >> +    +------------------------+------------------------+
> >> +    | emulator_id            | index                  |
> >> +    +------------------------+------------------------+
> >> +    | emulator_ctx                                    |
> >> +    ...
> >> +    +-------------------------------------------------+
> >> +
> >> +--------------------------------------------------------------------
> >> +Field            Description
> >> +------------     ---------------------------------------------------
> >> +emulator_id      0x00000000: Unknown (In the case of a legacy stream)
> >> +
> >> +                 0x00000001: Qemu Traditional
> >> +
> >> +                 0x00000002: Qemu Upstream
> >> +
> >> +                 0x00000003 - 0xFFFFFFFF: Reserved for future emulators.
> > Would it be useful for future proofing to carve out some space for a
> > per-emulator version field too?
> 
> What would that be useful for?  It is the emulators problem/fault if it
> can't read the blob it is given.
> 
> Superficially, I can see why the field would be nice for debugging
> purposes, but not all emulators will have a consistent version scheme,
> and we only install a single version of each emulator.  All I can see
> happening is libxl starting to guess about emulator/blob compatibility,
> which is absolutely not its place to do.

Good point (I also can't quite remember what I thought this for).

> > Otherwise LGTM.
> >
> > One thought, it might be useful (here or elsewhere) to have an explicit
> > overview of the expected control flow (as in the ownership of the fd,
> > and/or nesting of the layers as you prefer to think about it) between
> > libxc, libxl and the next layer (i.e. xl).
> 
> I will see what I can do, but the freeze is very imminent.

Indeed, don't let this distract you.