Re: [Qemu-devel] How to introduce bs->node_name ?

[Kevin Wolf]

Am 01.11.2013 um 16:12 hat Luiz Capitulino geschrieben:
> On Fri, 01 Nov 2013 08:59:20 -0600
> Eric Blake <address@hidden> wrote:
> 
> > On 11/01/2013 08:51 AM, Luiz Capitulino wrote:
> > > On Wed, 30 Oct 2013 13:25:35 -0600
> > > Eric Blake <address@hidden> wrote:
> > > 
> > >> On 10/30/2013 07:49 AM, Markus Armbruster wrote:
> > >>
> > >>>
> > >>> The first proposal is to add another parameter, say "id".  Users can
> > >>> then refer either to an arbitrary BDS by "id", or (for backward
> > >>> compatibility) to the root BDS by "device".  When the code sees
> > >>> "device", it'll look up the BB, then fetch its root BDS.
> > >>>
> > >>> CON: Existing parameter "device" becomes compatibility cruft.
> > >>>
> > >>> PRO: Clean and obvious semantics (in my opinion).
> > >>
> > >> I like this one as well.
> > > 
> > > Does this proposal makes "device" optional for existing commands? If it
> > > does then I'm afraid it breaks compatibility because if you don't
> > > specify a device you'll get an error today.
> > 
> > Changing from error to success is not backwards-incompatible.  Old
> > applications will ALWAYS supply device (because it used to be
> > mandatory).  That is, a management application that was intentionally
> > omitting 'device' and expecting an error is so unlikely to exist that we
> > can consider such a management app as buggy.
> 
> Doing such changes makes me nervous nevertheless. In my mind a stable
> API doesn't change. Of course that there might exceptions, but 99.9%
> of those exceptions should be bug fixes not deliberate API extensions.

Stable API means that whatever was working before keeps working. Nothing
less, but also nothing more.

If an extension that changes from error to success is breaking the API,
adding a new QMP command is breaking the API as well. Because sending an
unknown command means returning an error...

> A more compelling argument against it is the quality of the resulting
> command. I'm sure it's going to be anything but a simple, clean API.

I consider one command with an argument made optional a higher quality
API than having two different commands that do almost the same, except
that the older one has a mandatory argument where the newer one has an
optional argument.

Kevin