On 03/24/2010 07:25 AM, Paul Brook wrote:

...

I can't quite see what such a libqemu would buy us compared to straight QMP.

Talking QMP should be easy, provided you got a suitable JSON library.

I agree. My undesranding is this was one of the large motivations behind using JSON: It's a common protocol that already has convenient bindings in most languages. If it's hard[1] for third parties to bind QMP to their favourite language/framework then IMHO we've done it wrong.

You can't have convenient bindings to an RPC in C because it doesn't support dynamic dispatch. With most types of RPC, you have an IDL description and a code generator. But regardless of that, there are advantages to us providing a libqemu. The biggest one is that we can standardize transport implementations that include discovery mechanisms. If the core of libqemu provided an extensible transport interface, and a generic QMP request/completion mechanism, in a Python binding, you would never use the IDL generated wrappers but instead use dynamic dispatch to invoke arbitrary QMP requests. But the advantage is that if libvirt provided an API for a QMP transport encapsulated in their secure protocol, then provided the plumbed that API through their Python interface, you could use it for free in Python without having to reinvent the wheel. Regards, Anthony Liguori

Paul

[1] Hard compared to any other sane RPC mechanism. Some languages make everything hard :-)