On Fri, Jun 13, 2008 at 8:26 AM, Simon Horman [off-list ref] wrote:

On Thu, Jun 12, 2008 at 09:33:27PM +0200, Julius Volz wrote:

quoted

Ok, my first impression is that genetlink is aimed at being simple to
use (and has a nice howto).

So we'll work on a genetlink interface and some of the other v6 patch
issues and then post again in a while. Thanks for the feedback!

Horms: ping if you're interested or have some good ideas for this.

Julius: pong

The main two problems that I see in the existing interface are
a) lack of extendibility (which is why we are here) and;
b) non-idempotent actions, especially adding and deleting
  real servers, which mean that user-space programs that
  manipulate ipvsadm have have extra (racy) logic.
  (ok, perhaps that is more a pet peeve than a problem).

Ok, so we probably won't focus on b) as a priority right now, unless
it happens as a side-effect.

I don't really have any concrete ideas about what a better
interface would look like. But I am more than happy to hash our ideas.

Good! At the moment I'm looking at various netlink docs and figuring
out how things generally work. I think netlink probably adds a lot of
complexity over the previous sockopt interface, but I hope it's worth
it.

As for compatibility and extensibility, how is that best achieved with
netlink? I've seen some examples copy whole C structs into netlink
datagrams, but that is obviously what we don't want anymore. So the
way to go seems to be to transfer each struct field as a separate
netlink attribute, right?

Julius

-- 
Google Switzerland GmbH