On Thu, 04 Mar 2010 22:06:42 +0100 Markus Armbruster <arm...@redhat.com> wrote:
> Luiz Capitulino <lcapitul...@redhat.com> writes: > > > On Thu, 4 Mar 2010 16:56:49 +0100 > > Markus Armbruster <arm...@redhat.com> wrote: > > > >> While fully converted handlers are not supposed to print anything when > >> running in a QMP monitor, they are free to print in a human monitor. > > > > I disagree. > > > > One of the key decisions behind the new Monitor design is that handlers > > are part of common code. User printing is output-specific and should not > > be done by handlers. > > > > One of the problems with this layer violation is that the consumer of this > > data might not be what you have assumed. For example, it has been suggested > > that we could move the qemu shell out of qemu in the future. In this case, > > this kind of information _could_ be useful. > > How do you intend to solve the problem of printing help then? Doesn't a shallow conversion takes care of it for now? > The perfect is the enemy of the good. > > > Another important problem is that free printing like this is error-prone > > as it's not at all clear what could be printed and there's no way to catch > > mistakes mechanically. Not to mention that this is for sure going to used > > by those who want to ignore QMP completely. > > This paragraph I don't get. If something that should be available under qmp is printed by using monitor_printf(), we loose big time.