Re: [Qemu-devel] [RFC][PATCH 0/2] Allow cache settings for block devices to be changed at runtime.

[Anthony Liguori <anthony@codemonkey.ws>]

On 03/01/2011 08:22 AM, Kevin Wolf wrote:
>> Certainly good questions, but let me suggest not taking an HMP command
>> and not a QMP commans because of interface concerns.
>>
>> My goal for 0.15 is to convert HMP to be implemented in terms of QMP.
>> To do that, a bunch of new QMP commands are needed.  They all won't be
>> perfect but i'd rather support a bad QMP command forever than to
>> continue to/ have people rely on HMP.
>>      
> Okay, makes sense. So we should reject patches that add new HMP commands
> without adding a QMP counterpart.
>    

Definitely.  We essentially are supporting HMP today just as much as QMP 
but HMP is much harder to support (no standard way to interpret 
input/output/errors).

>>>> I agree that the guest should control the
>>>> emulated drive cache at runtime and we probably don't want to allow
>>>> toggling that from the host - it could be dangerous :).
>>>>          
>>> Good point. That's a NACK for this patch as long as we haven't separated
>>> WCE from the host cache setting.
>>>        
> Doesn't make a difference for this one, though, because it's NACKed anyway.
>
> Kevin
>
> PS: Anthony, is there a specific reason why you started sending HTML emails?
>    

Because I was stuck using my phone because of bad hotel wireless :-/

Regards,

Anthony Liguori