> The support for the "key" dladm(1M) parameter to refer to an aggregation > is there strictly for backward compatibility. The preferred method to > refer to an aggregation link is by name, and it's awkward in the > documentation to always say that one can either use an integer key or a > link name. Such statements are numerous since there are a number of > aggregation dladm subcommands. > > I'm thinking that it would be appropriate to remove all references to > keys from the documentation, and only leave support for it in the command > as an undocumented thing to not break existing scripts and trained > fingers... Do others see this as problematic?
It makes me a bit uncomfortable to completely eliminate something from the manpage that is still a committed part of the interface, though I agree it's clunky to lug it around throughout the manpage. Maybe on first reference we could mention that a key is also supported (though use of link name should be preferred) and point the reader to a NOTES section that provides additional background on how link names are constructed when keys are used and other details like that? -- meem
