On Tue, Mar 18, 2008 at 08:01:01PM -0700, Bart Smaalders wrote: > Henry B. Hotz wrote: > > >I find it really > >difficult to invest time in learning single-platform technologies. > > That makes it difficult to do innovation, since we need to convince > other OSes to use our technology before you will use it :-).
This applies to other OSes too though. If some flavor of Linux wants to innovate, do the users of the other distros resent it? For a while, maybe, but if the innovation is good enough it may come to be accepted. > So far we've succeeded w/ DTrace and ZFS; while many OS developers like SMF > because it solves a bunch of hard problems, many system admins have > complained loudly. As a former sysadmin and current OS developer, I really like SMF. Yes, there's a few places where it needs improvement, but those are nothing compared to the problems it solves. And yes, it solves real problems (at the very least the idea of process contracts and service restarters does; the configuration repository is what I see complains about, but frankly, I like that too). I think a remote access protocol to SMF would probably make things easier. Perl5 and Python bindings for it would go a long way too. Anyone can add those, but it might be better if *we* did it, even if there's no identified customer (don't underestimate the value of CPAN module availability for any given OS technology). > Personally, I now hate dealing w/ pre-S10 systems, since i have to remember > how to enable/disable each sub-system separately... and I have to redo that > every time I update the OS on the box. > > W/ SMF, I always know where my log files are, how to do basic admin and > (above all), I can now log into my machine when the building yp servers > are completely > fubar'd. All what Bart said.
