It seems to me Linux folks would prefer ethX naming, and Solaris folks don't have a problem with the current naming (although trying to remember that nge and nxge are not even remotely related is kind of archaic). In that respect, it would almost seem worthwhile to have an option on a clean Indiana install:
+ Network Device Naming + (*) Linux Style (default) ( ) Solaris Style I guess my opinion on netX is a different from a couple of days ago when it seemed like a good idea. Now it just seems like it would confuse both camps and not make anybody happy. Heck, we're using Clearview already to name NICs after the VLAN shortname we use internally. So it doesn't really matter for our environment if its netX. -J On Wed, Apr 23, 2008 at 4:06 PM, Peter Tribble <peter.tribble at gmail.com> wrote: > On Wed, Apr 23, 2008 at 4:29 AM, Peter Memishian > > <peter.memishian at sun.com> wrote: > > > > > Folks, > > > > As some of you may be aware, Clearview Nemo Unification and Vanity Naming > > integrated into Nevada build 83. Among other advantages, the "vanity > > naming" feature allows us to finally[1] move away from the current chipset > > alphabet soup we have for network interface names, and instead standardize > > on simple names like net0/net1. For more background, see: > > > > http://www.opensolaris.org/os/project/clearview/uv/ > > http://www.opensolaris.org/os/project/clearview/uv/howto > > > > In Nevada, backward compatibility concerns have stopped us from changing > > the default naming convention, so one is still stuck with unintelligible > > names like e1000g2. > > I'm clearly in the minority here, but in what way is something > anonymous like "net0" any more intelligible than "e1000g2"? > > I see no advantage here. I've been lumbered with this sort of > random naming scheme elsewhere and it's a real PITA. > > > > However, Indiana's experimental nature seems to make > > it an ideal vehicle for making such a change, and its target userbase > > seems suited for benefitting from such a change (e.g., networking > > procedures aimed at new adoptees could simply assume the first network > > interface is net0, rather than using placeholder interface names). > > So what determines the numbering scheme in a system with multiple > interfaces (in particular, the order in which they're enumerated)? If I > primarily use wireless, does that end up being net0? > > It seems to me that there's a fundamental weakness that needs to be > addressed. If users are doing low level admin stuff that requires them > to know the name of the device - even if that name is net0, then we've > lost. On many systems they're still going to have to work out what the > name of their network interface is (which of the net0 ... netX should > they use?). And once you're down at that level, "friendly" names > have no benefit. > > > > Thoughts? > > > > PS. Please note that device information is still available -- e.g., > > post-build-83, "dladm show-phys" will show the underlying device > > (e.g. e1000g2) associated with each link. Ongoing work (such as > > NWAM's GUI) will likely provide detailed hardware information about > > each network interface. > > But why force users to go through those extra hoops? And why can't the > admin interfaces be responsible for simplifying the user's life, rather than > using underlying device names? > > > > [1] This happened forever ago for other hardware like disks -- can one > > imagine having /dev/seagate0 and /dev/maxtor1? Yikes. > > And you want to go back to sd0, sd1, etc? > > -- > -Peter Tribble > http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/ > > > _______________________________________________ > indiana-discuss mailing list > indiana-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/indiana-discuss >
