On Mon, 2009-03-23 at 14:14 -0400, Girish Moodalbail wrote: > > Also along the same lines, the -f option for the *-prop subcommands > > shouldn't be necessary. > > > The idea here is, instead of having separate tunable for v4 or v6 (read > ip_forwarding or ip6_forwarding) we could specify the -f flag for 'ipadm > set-prop' to identify which 'ill' the property needs to be specified.
The argument that is being made is that "ill" is an implementation detail that has leaked into the administrative model and confused people who are familiar with other operating systems (e.g. BSD and Linux). Are you saying that there will be a single IP interface, but that there will be multiple properties with the same name associated with that interface? That doesn't seem desirable. How do these properties get disambiguated in the show-prop output for a given interface? > Also for example the 'ip_fragment_timeout' which needs different timeout > value for v4's and v6's interfaces can be easily achieved using the '-f' > flag instead of defining two different variables. I don't see the problem with different variables. The variables are specific to a protocol operating over an IP interface. The model that I'd like to see is one where there are multiple protocols operating over a single IP interface. Some interface properties could affect the operation of one protocol over that interface (e.g. ipv4-forwarding and ipv6-forwarding, and others could affect any protocol over that interface. This doesn't seem problematic to me, at least not relative to the aberration that is the current Solaris IP interface namespace. -Seb
