[OpenIndiana-discuss] I figured this deserved a separate thread

[Kevin J. Woolley]

On Thursday, 4 November, 2010 16:02, "Frank Middleton" <f.middleton at apogeect.com> said:

> Well, it could be hardware too. But for whatever reason - memory leaks,

You'd be surprised at the hardware I run it on -- old and grungy are the orders of the day for a great many of my systems.  My primary point is that the success of maintaining/administering any serious OS (of which Linux and at least two of the BSDs are, and Windows arguably is ;) ) lies largely in areas outside the realm of the OS -- the skill and experience of the admin, the non-OS software that's run on the machines, and so on.

> other OSs have improved in the last year or so but updating them
> is so difficult (or impossible without re-installing) in comparison.

That depends strongly on the distribution.  FreeBSD has a *very* strong upgrade path, save for major version changes.  CentOS and other enterprise-grade Linuxes also upgrade well among minor releases, but need to be re-installed to go to the next major release.  Those release cycles are suitably long that I'd argue the net effect is negligible.

Fedora, most Ubuntu (minus LTS), and many other Linux distros have much shorter release cycles and don't make any claims that they upgrade well among actual releases (as opposed to repo updates, etc.).  Saying that Linux as a whole is deficient because of these cases is a bit like saying screwdrivers suck because they don't hammer nails well.  Different tools (and distros) for different jobs.

>> Solaris is a great OS. It doesn't need people pissing on everyone
>> else's to be good.
> 
> Absolutely. But for it to take off it has to be better than good. My point
> only was that for reliability, scalability, backwards compatibility, and (I
> forgot to mention) upgradeability, it simply can't be beat. As just another

For what it's worth I don't believe I've ever upgraded a Solaris box across major versions, so I don't have any direct experience there.  I agree that Solaris' ability to do this is great and other OSs could certainly improve in this area -- so why not leverage this fairly unique trait of Solaris and see how it can be made even better?  No one has *ever* been interested in using product/project $foo because they're told their current product/project sucks.  I don't say this as a disgruntled Linux admin (I also admin Solaris 8, 9, 10, OI, two versions of AIX, three of HP/UX, two of FreeBSD, four of Windows, and so on and so forth).  I say this as someone who is quite honestly sick and tired of "$foo sucks, $myfave rocks" being a dominant attitude all over the tech community.

> Gnome based desktop OS, maybe it doesn't have any particular advantage,
> although IMO the pkg system beats apt and yum et al for ease of use and
> with ZFS it is so easy to roll back. These aren't features you can easily put

And that's great -- we've all got our favourites, and Solaris and OI contain some of mine.  Should the focus be on improving what we're working on or locker-room talk about what others work on?

> in a list along with a better analog clock :-) and have their roots in  an
> architecture that's hard to replicate. If we can't compare it to other OSs
> aren't we lost before we start?

Useful comparisons are great.  "$foo works better than $bar because of X" has good signal.  "$foo works better than $bar because of X and here's how we can make $foo even better" has even better signal.  "$bar is a steaming heap that can't possibly touch $foo" is almost entirely noise.  (I'm exaggerating for effect here, so let's leave out the argument that's not what you said.)

The Solaris community (spanning commercial offerings, OpenSolaris, Illumos, and OI) has a *LOT* of knowledge that's not easily gained and requires experience to obtain.  My opinion is that a very high percentage (80% or so is my current gut feel, subject to change by the hour/day/phase of the moon) of the problem of getting people on board with OI is making that knowledge easier to obtain.  As with anything else in life, if there's a complicated but wonderful thing in front of you it doesn't mean beans unless you can figure it out.

Sure, there's a technical component there as well.  OI has rough spots that need to be smoothed over, and in the medium to long term Solaris has to continue to evolve.  That is also crucially important.  I just don't see what taking potshots at other peoples' projects contributes to either of these goals.

Cheers,

kjw