From: Jiri Pirko <j...@resnulli.us> Date: Tue, 15 Mar 2016 09:24:17 +0100
> Tue, Mar 15, 2016 at 09:08:44AM CET, ro...@cumulusnetworks.com wrote: >>Apart from the usability concern i have described earlier, this just seems an >>overkill ...having to redefine every attribute. > > I don't hear the usability concern. User knows how to compose/parse > attrs, no problem there. But if we compartmentalize the most common use case into the header itself, then the user doesn't even have to deal with attributes at all. This is what we've (intelligently, not stupidly) done in the past. When we build a new facility, we put the most fundamental elements into the message header structure. There is nothing wrong with this at all. If we run out of bits for the filter mask, we can (intelligently, not stupidly) add a new attribute for that. This is nothing in the realm of "good netlink design" that says that we should use attributes for everything. In fact I would say that netlink attribute diahrrea should be avoided. Therefore I am more than happy to see the filter mask stay in the message header, and I will apply changes from Roopa which implement it this way. I've read your arguments, I understand your position and your reasoning, but I simply disagree with you. Thanks.
