> Like I said above, my local understanding about
> David's comment is that 
> he meant OpenSolaris is a Linux replacement /for the
> LAMPS niche/ .   
> Once again, OpenSolaris is about creating a superior
> OS for targeted 
> niches, of which LAMPS is one.  To win in such
> niches, you do need to 
> provide functionality equivalent or better than your
> competitors.

OpenSolaris will never make it to the top of the food chain because it has 
severe architectural issues, starting with the software management subsystem, 
continuing with breaking compatibility with Solaris, and causing tremendous 
engineering and software development effort for third party ISVs, of which I am 
one.

And: OpenSolaris will never be able to compete with Solaris 10 enhanced with 
Oracle, PHP and Apache.  Not with all the fancy freeware, not with all the 
gizmos and GUIs, it won't stand a chance against Solaris 10.  It's too unstable 
to be useful for any kind of system engineering on top of it.

Unstable in terms of the software, unstable in terms of architecture. Severely 
lacking in enterprise features (more on that below).

All that was really needed was Solaris 10 with tons of prepackaged and ready to 
go freeware, integrated into the OS and kept "fresh", and more bug fixes, not a 
complete reengineering effort. That was a strategic mistake.

> To this end, the common GNU utils found in Linux are
> now a standard part 
> of OpenSolaris, in /usr/gnu , which is now in the
> default path. This 
> replaces the old way of prepending 'g' to the gnu
> tools, which used to 
> be stored in /usr/sfw.  The old SYSV utils haven't
> moved from /usr/bin , 
> and aren't going to be moved or replaced.   BASH is
> now the default 
> shell for root, and I know this is a big thing for
> many people, as Bash 
> is not a 100% Bourne Shell replacement.  /sbin/sh is
> still there, as is 
> the fully-compliant /bin/ksh.  

Yes yes, this is all known to me. And to have that changed to the way it was, 
it causes system engineering overhead for companies and ISVs.

That costs money. So by making these decisions, someone cost me money. To whom 
and where can I send the bill, please?

> You're mis-informed, as I've stated above.

That is EXTREMELY UNLIKELY, as I track changes in (Open)Solaris down to the 
file level.
Almost daily.

> GNU tools
> are NOT being 
> modified,

"David Comay      2008-05-21 22:57:00 UTC

We're going to be going through these utilities, one by one, and working on
determining which version we wish to use as a base and which changes are
necessary then going forward."

http://defect.opensolaris.org/bz/show_bug.cgi?id=576

> There is indeed a failure here, and it's one of
> communication:  Sun 
> certainly needs to be clearer about exactly what
> OpenSolaris is 
> targeting, and how it is going to get there. _I_
> certainly could use 
> better information.  And, yes, I do wish it were
> easier for outside 
> folks to contribute (and feel like it's worth-while
> to contribute) more 
> code.

Except that people which OpenSolaris is targeted at are extremely unlikely to 
contribute any code, because they simply do not have the necessary technical 
expertise - an expertise of a UNIX kernel engineer. Wrong crowd, and a gross 
miscalculation. Or perhaps, nobody even stopped and considered it, which is 
more likely.

Those people aren't kernel or system engineers, they are "me toos".

> How does it NOT feel like Solaris anymore?  Because
> your default path is 
> different?  Because we've dumped the ancient (and
> frankly sucky) svr4 
> packaging system?

ABSOLUTELY!
Not only was SVR4 packaging ditched, instead of being INCREMENTALLY IMPROVED 
the Toyota way, the designers of IPS thought themselves extremely clever by not 
providing pre- and post install and remove mechanisms.

As a result, SMF is now being overloaded with the equivalents of pre and 
postinstall and remove scripts, because people are basically forced to use SMF 
as a backdoor for automation and reaching CMM level 2 and above. Nice.

Because of course, if Shawn Walker & Co. can't imagine that someone would want 
to package preconfigured configuration files, or have Oracle configure herself 
on the fly during installation, it doesn't exist.

> Because more GNU (and other
> freeware) packages are 
> now included as part of the default install, instead
> of on the Software 
> Companion CD?  Because the hideous piece of crap that
> was CDE is being 
> dumped overboard 10 years late?

I couldn't care less what GUI du jour is used or dumped, because a properly 
engineered system doesn't even need a GUI. When you have tens of thousands of 
systems in a lights out management environment, the desktop GUI concept is 
completely useless, and that is the case here with me.

Solaris 10 needed more software, yes, integrated into the operating system, so 
that people like myself (and others) didn't have to spend sleepless nights 
integrating Postfix for instance.  That alone cost us about $18,000 USD.

We needed incremental improvement, more built-in software, and bugfixes, not a 
completely different solution.

> Or that LiveUpgrade
> is replaced with 
> beadm ?

Couldn't care less about ad hoc methods like live upgrade; that is something a 
Linux kid would do, upgrade a production system ad hoc, or even worse, right 
off of some repository on the internet.

That's not how that's done, at least not in mission critical environments.

What I do care about, on the other hand, and care about greatly, is 
JumpStart(TM) + Flash(TM). OpenSolaris has nothing that can even touch that 
technology, and that alone speaks volumes about it. Automated installer 
technology in OpenSolaris is ad hoc at best. No concept of engineered system 
images, they would have us pull and install packages, one-by-one, off of some 
repository. I write these things and I shake my head, that's simply 
unbelievable. OpenSolaris will destroy and consume itself.

But hey, Linux has it, so we have to have it, too! Yeah, why be a technology 
visionary and a leader, when we can go the "me too" route and try to emulate an 
inferior solution?

> The changes you see in OpenSolaris vs Solaris 10 are
> large, but then 
> again, so were the changes between Solaris 9 and 10.
>  Remember SMF?

I REMEMBER SMF very well; after all, I've integrated my own SMF changes into 
SX:CE; and had it not been for deep, deep knowledge of Liane Praza and Casper 
Dik and David Bustos to explain the inner workings of SMF, I'd still be stuck. 
And if I had a Dollar/Franc/Euro for every time I was told "you can't do that", 
or "that's planned, but not there yet", or "we never thought of that"...

SMF is actually great, but it requires lots and lots of study and effort to 
comprehend.
  
> BAC?

RBAC is too complex for what it was meant to do, and will never dethrone 
sudo(1M), because sudo(1M) is "good enough", does the job perfectly, and is 
multi-platform, while RBAC is non-standard, found only in Solaris and nowhere 
else. That should not be the case for a piece of software as critical as that.

> Can 
> u tell me what is wrong in the new OpenSolaris vs
> Solaris 10? I'd be 
> curious to see what you dislike.

I dislike that fact that default is GNU.
I dislike the fact that root's shell is /bin/bash.
I dislike the fact that SVR4 packaging tools haven't been enhanced sufficiently 
to offer the capabilty that hp's SD-UX and sgi's inst(1M) offered for the past 
20 years or so.

I dislike the fact that if it was decided to implement a new software 
management subsystem, that subsystem is NOT sgi's inst(1M), which can do 
everything IPS can do, and a hell of a lot more. The inst(1M) software 
management subsystem understands SVR4 packages, and it is still more advanced 
than IPS, AND it can be had for peanuts because sgi IRIX can be shelved, so I 
extremely dislike that it wasn't bought and made into the software management 
subsystem in Solaris.

I dislike the fact that IPS is written in Python and that it is slow.

I dislike the fact that there is no equivalent of Flash(TM).

I dislike the fact that the hundreds of System V packages we toiled so hard for 
are now worthless, all that automation - worthless, all that system engineering 
- worthless, because someone thought that Solaris should be "hip".

Well, it already is hip. And cool. And much cooler than any Linux will ever 
come close to. But because of all the "me toos", sysadmins and system engineers 
and ISVs are now getting screwed.

That's simply "phenomenal".

Go ahead, trade us in for a bunch of GNU/Linux kids, "me toos". Let's see if 
they can ehance your kernel and utilities, and fix your memory leaks, core 
dumps, and mutex locks. 
Let's see if they can come up with the next HA clustering software, next ZFS, 
next SMF, next inst(1M).
Let's see it.

> And I'm still waiting on a unified
> management interface for 
> everything.

Interfaces are useless without solid architecture and sufficient kernel and 
system engineering expertise.
-- 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to