Darren J Moffat wrote:
> I think both phases of this project are a very bad idea and I do not
> approve of the strategy.
>
> The Solaris security team did a similar analysis for 'pfexec' vs
> 'sudo' for almost identical reasons as this case. Like this case we
> planed to initially ship a symlink pfexec -> sudo and later enhance
> pfexec with the functionality that sudo had. In the end we choose
> not to do that because it as more engineering effort, it would never
> be perfect and there really was no harm in having the real sudo
> installed on the system. So that strategy was abandoned and now
> Solaris/OpenSolaris has the real sudo.
>
> I strongly suggest that this case be withdrawn and a new case be filed
> to just ship top.
>
> This is is also misnamed there are no prstat enhancements in this case
> just a confusing symlink from prstat -> top.
>
> I'm think that prstat could do with some enhancements but this case
> didn't actually provide any.
As for me, I'd agree with phase 2, if phase 2 provides near-enough
approximation of functionality that users are unlikely to notice the
difference. But phase 1 seems more and more like a terrible idea.
I think the problem at hand is that top may lack support for some
Solaris features (zones, projects), and uses direct kmem access instead
of nicer methods such as /proc.
So that begs the question: which problem is harder to solve: give prstat
top-like functionality (enough so users will not notice the difference),
or give top enhancements to make it competitive with prstat?
-- Garrett
