> Yes. This is another reason to switch to Linux. On this platform most
> applcations for Unix-like systems will be supported. And on Solaris when you
> need some program not from top 100 list ( for example, fsvs, a lot of user
> applications), you more often should dance with tambourines and dig into
> application code to make it work...

That is true because the development of those apps is Linux-centric and
making them work on anything else is an afterthought.  Therefore, the
greater availability of apps is also not free of trade-offs.  Such a
development approach will have more stability problems, as well as
tending to lock you into the platform.  The open-source-must-rule
GPL-is-the-best-license Linux-is-all-that-matters crowd always
considers themselves morally superior to Microsoft, but if the
effect either way is lock-in, I personally think that's still a problem.

If their engineering discipline and compliance with platform-neutral
standards were better, I'd say that at least their ideology didn't cause
other problems, although I'd still find it impractical - what do I care
if someone wants to add a little functionality to open-source software
and then call it a proprietary appliance?  My _printer_ (Ricoh CL3500N)
runs some flavor or another of BSD rather than Liux for that reason,
that they didn't want to fool with how much of their proprietary bits
they might have to open.  I admit I wish they had, but I don't really need
them to have had.   The problem I have with them is that they're horrible
about providing advanced info on how to do stuff like upload
fonts, forms, and the like to an internal disk, since they only provide
a tool for that which runs on Windows.  All they'd have to have done
is released some info on how to _use_ the printer fully, that's little
different from what it would be on similar printers from a competitor.
That's also another example of why I think being open about standards,
formats, protocols, and communication is even more important that
the source code being open.  Most people just want to _use_ the darn
thing, or maybe sometimes write a tool that will interact with it
in reasonable ways, for platforms that don't already have such a tool.
I'd enjoy OS hacking if I really had time to get into it, but most
people would rather be able to take the OS for granted.

> I would consider Solaris/OpenSolaris for new
> deployments if it had free updates. Otherwise for me
> it is unnecessary waste of money for support (at
> least, at campus).

One can still just upgrade once or twice a year for free;
there are overall updates roughly that often.
(give or take whether the license for Solaris allows you
to use it in the capacity that you want to use it for)

OpenSolaris upgrades to new builds (as opposed to
fixes to an existing build) are also free, right?

A support contract gives you patches to an existing
build in addition to overall updates.  It also gives
you some assurance that someone will at least acknowledge,
and perhaps even constructively respond to, any bugs
you may encounter.

On Solaris, if you plan for it by having enough space
(easier with zfs root), you can use Live Upgrade,
such that you can (a) create a new boot environment
from the existing one, (b) update the new boot environment
with either patches or an upgrade, and (c) reboot into
the new environment, still having the option to revert to
the old environment.  Not sure the present status, but
a somewhat similar capability is said to be in the works
for OpenSolaris as well, last I heard.

Patches (or individual package updates) to an existing
build are mostly an issue for systems one wants to keep
not so much bleeding edge as stable and secure.  Security
alone could be achieved with upgrades (free) instead,
except that they may not be timely enough.

So it depends on what model you want, and what sort
of support you consider sufficient.  With most Linux distros,
even if you didn't have to pay for anything, you'd effectively
have to do more work yourself (in terms of tracking issues and
determining what updates were stable enough to use, even if
not in terms of building anything) than you would with a good
support contract.

Nothing is ever truly free; consider the cost of having additional
bodies of your own to handle keeping track of everything one
has to with a distro where both the distro and all updates are
free.

I don't know about BSDs in general, but the sort-of BSD derived
Mac OS X XNU kernel (Mach + in-kernel BSD code) has MacPorts,
which is a bunch of common Linux+BSD applications in source form, using
the "ports" packaging approach originating on (some flavor of) BSD.
Problem is, it's source-based, so if there is some combination of
builds and/or build options that doesn't work or doesn't satisfy
dependencies, you're out of date on some things until that gets
resolved (and building everything yourself from source is _slow_,
even when it's painless).  Other than that (wanting the option
of binary apps and updates, in known-to-work-together combinations),
personally I'd prefer FreeBSD to Linux, depending on what my needs were
(i.e. driver and apps support), due to BSD sharing more recognizable
lineage with Unix, zfs, and that I'd have less problems with the
ideology that comes with FreeBSD than that which Linux tends
to be burdened with.   (I like the BSD license better than the
ideological baggage of the GPL, and I think the BSD developers
come from a more disciplined tradition - more like engineers than
like a bunch of very talented but largely self-taught people that
are perhaps short on concepts like standards, interfaces, layering,
and the like.)  And last I heard, Linux SystemTap was nowhere
near what DTrace can do, although if one takes into account
whatever limitations of DTrace on FreeBSD may be, perhaps
the difference isn't quite as great.

At one time it was the case that you'd pay more for RedHat
support than you do for Solaris.  Right now, it doesn't look that
way to me, but the situation may not necessarily have finished
changing yet.

Companies that have to pay salaries can't give away both the
software _and_ the maintenance, they've pretty much got to
charge for one or the other.

Solaris is meant to be stable; some things get backported, but
not everything.  OpenSolaris is (or at any rate has been, I
don't know to what degree anyone is 100% comfortable with
what Oracle might do) where the development takes place.
It's much more leading edge than Solaris, but you definitely
have to be prepared to have a test system _or_ to wait with
upgrading to the latest build until you hear how it's doing
before throwing it willy-nilly onto "production" systems.
Not that it can't be used on production systems, some places
do; but it would require great care and wouldn't be appropriate
for all sorts of production systems.

Whatever you do, expect to re-evaluate in maybe six months to
a year.  (IMO) Sun mostly seemed to "get it" when it comes to the
value of being in an educational setting, namely that if they interest
people there, they're likely to have a foot in the door with the next
generation of paying customers.  OTOH, Sun did have some trouble
making a profit.  Whether Oracle figures out how to achieve a
balance that puts (Open)Solaris where it will be likeliest to be
adopted and flourish, while still being able to have it (and
associated hardware perhaps)  make some profit, I think remains
to be seen, and needs periodic re-evaluation.  Right now, I think
a lot of folks may suspect that Sun gave away too much, and
Oracle gives away too little (in the free as in free beer sense,
since AFAIK Oracle hasn't changed anything in terms of the
free-as-in-open-source sense, except for being a lot more
closed when talking about future developments).

I _know_ that there are benefits to be had by separating
specification and standards from implementation; but a lot
of those benefits are more long-term than a lot of people care
about.

I _think_ there are benefits to be had by having a _loose_and_flexible_
relationship between the usage model and how development and
maintenance is paid for, but I admit I'd have a harder time arguing that one.
-- 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to