On Wed, Dec 21, 2022 at 10:21:16AM -0700, Theo de Raadt wrote: > Claudio Jeker <cje...@diehard.n-r-g.com> wrote: > > > In the old world only netstat could show the routing table. I think this > > is still the case in FreeBSD for example. We added route show at some > > point but the documentation was not shared between the manpages. > > I agree that it is annoying to open up the netstat man page to find the > > flags shown by route show. For this we added: > > > > Print out the routing table, in a fashion similar to > > "netstat -r". The output is documented in more detail > > towards the end of the netstat(1) manual. > > > > To the route manpage when describing route show. Not sure if that is > > enough or if we should duplicate tables (whith the usual sync problem). > > This is all true. > > So netstat in those days was more of a "kvm reader" program, and as such > racy. Nowadays both route & netstat programs's subcommands are a mix of > sysctl readers and route socket askers/listeners, and thus they have > better atomicity or at least other types of truth. > > The sub-command extensions in route are better designed, mostly because > they are newer and were built in an era where kernels maintained more than > a handful of routes. I think we want to lean people towards using route, > instead of netstat. > > So I think route.8 is where we should try to have complete documentation, > and once that is done we should change Xr's and other documentation to > point at route.8 instead of netstat.8
The show code of netstat and route should be in sync. I did that long time ago. The kvm bits are only used in netstat and may be broken again. Did not try lately. Still I think 'route show' is more obvious and has a few extra arguments over netstat -rn. -- :wq Claudio
