Re: [Qemu-devel] [PATCH v2 2/9] qapi: add socket address types

[Luiz Capitulino <lcapitulino@redhat.com>]

On Tue, 02 Oct 2012 05:39:52 -0600
Eric Blake <eblake@redhat.com> wrote:

> On 10/02/2012 03:00 AM, Paolo Bonzini wrote:
> > Il 02/10/2012 01:56, Eric Blake ha scritto:
> >> On 10/01/2012 08:52 AM, Paolo Bonzini wrote:
> >>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> >>> ---
> 
> >>> +#
> >>> +# @port: port part of the address, or lowest port if @to is present
> >>> +#
> >>> +# @to: highest port to try
> >>> +#
> 
> >>> +{ 'type': 'IPSocketAddress',
> >>> +  'data': {
> >>> +    'host': 'str',
> >>> +    'port': 'str',
> >>> +    '*to': 'uint16',
> >>
> >> I still think it's a bit weird to have:
> >>
> >> 'host':'localhost', 'port':'1000', 'to':1001
> >>
> >> for a port range, all because of the possibility of named ports; should
> >> '*to' be a 'str' if only for symmetry in the output?  But it's
> >> bike-shedding, so I'll live with whatever works (that is, I'm not
> >> requesting a v3 on this patch).
> > 
> > Would it be better if I changed 'to' to 'count'?
> 
> That does look a little better:
> 
> 'host':'localhost', 'port':'1000', 'count':2
> 
> for the 2-port range 1000-1001.  But it's all the same information, so
> I'm not strongly tied to any particular representation, as long as
> libvirt can parse it when querying and produce it when starting NBD.

Wouldn't it be cleaner to pass a list of port numbers? We could have:

 *port-list: [ 'int' ]
 *service: 'str'

Or, we could make this an union.