It seems *awfully* strange to me that we feel obliged to integrate a set
of command line utilities which are designed to provide raw access to a
hardware bus to the command line.
While I suppose some might argue that there are some precedents
(pcitool) -- I think those precedents are somewhat limited, and
generally have little chance for destructiveness. And, I don't believe
I'm alone (but I may be in the minority) in my belief that generally
this is the kind of detail users *don't* want to see. (If you look at
other shrink wrapped OS', I think you'll find that neither Windows nor
MacOS offer this kind of direct access to users.)
The whole idea that we need to do this at all makes me believe that
there is some other set of larger architectural gaps in the system. (Or
is it the case that there aren't, and this is another case of "it ships
on Linux so it shall ship on Solaris"? I hope not, because apart from
some very specific diagnostic needs, I believe that Linux' inclusion of
this in certain distributions is also a sign of architectural
incompleteness in Linux itself.)
While I agree that diagnostic utilities can be useful in trouble
shooting certain problems (probably most often in development of new
device drivers, etc.) I hope that we don't believe that these kinds of
utilities are the sorts of things that average, or even advanced,
Solaris users or developers are going to be expected to use.
So with that in mind, I think I agree with Jim. The folks using these
utilities should be smart enough to figure out to run the commands as
root, without needing preconfigured RBAC. Just because a user logs in
to the console, I don't think he should have access to the raw busses of
the console devices (including USB, SCSI, and any 8042 style busses.)
I don't think the failure to "automatically" support these utilities
would represent any "feature gap" on Solaris. (But then again, I don't
think a failure to "ship" these utilities represents any such feature
gap either.)
-- Garrett
