SUMMARY OF REPLIES TO MY REQUEST FOR SOLARIS INFO.
1) What is the differences between solaris and sun/os?
------------------------------------------------------
A> Solaris 1.0 is SunOS 4.1.1 with OpenWindows 2.0 and ONC (RPC/NFS/etc)
A> Solaris 1.0.1 is SunOS 4.1.2 with OpenWindows 2.0 and 3.0 and ONC.
A> Solaris 1.0? is SunOS 4.1.3 with OpenWindows (2.0 ? and) 3.0 and ONC.
A> Solaris 2.0 is SunOS 5.0 with OpenWindows 3.0.1 and all the rest....
B> Solaris 2.0 is SVR4 with a very clean internal implementation.
B> It can do MP.
B>
B> SunOS did not have SMP. It was BSD based. It has grown ugly
B> over years of hacking.
B>
B> Solaris 2.0 will likely be _more_ efficient after tuning.
C> Solaris is based on System V Release 4, not on Berkeley Unix. Many,
C> many things have changed (ps -ef instead of ps -ax, etc). It has a
C> Berkeley environment, but it's not the easiest to work with.
D> Solaris 1.0 is SunOS 4.1.x bundled with a bunch of other stuff like
D> OpenWindows v. 3.0 and Deskset 3.0.
D>
D> Solaris 2.0 is SunOS 5.0 (roughly sVr4.0.4 with MP extensions) bundled
D> with the same stuff as above.
D>
D> Solaris 2.1 is as above, but more stable and with symetric
D> multiprocessing for their MP machines (Solbourne still blows them away
D> in this area, however).
E> 'Solaris' means Solaris v2 which is essentially another name for SunOS v5,
E> which is based on System VR4, with compatibility-to-BSD additions.
E>
E> 'SunOS' means SunOS v4.x, also known as Solaris v1.x, which is BSD based.
F> Solaris is an implementation of AT&T/USL System V release 4, as opposed to
F> SunOS which is a mixture of SysVr3 and BSD - sort of a "best-of-both-worlds"
F> OS. SysVr4 is reputed to be very nice, and includes many features previously
F> only found in BSD.
G> Lots. sun/os is primarily BSD with optional System V enhancements, whilst
G> solaris is primarily System V (Release 4) with optional BSD enhancements.
G>
G> HP/UX is mostly system V release 3, whilst AIX is like nothing on this
G> planet. (see personal opinion below)
H> Solaris 2.x/SunOS 5.x is AT&T System V Release 4.
I> Almost everything. Except neither comes with Motif :-)
2) Are they claiming complete backward compatibility?
-----------------------------------------------------
A> SunOS 5.0 is a major step. It gives you binary compatibility if you application
A> is well-behaved. If you are on software support, Sun should have sent you a
A> package called "Solaris Migration Kit". You should definitely read it.
A>
A> In short terms, your program must be dynamically linked, it must not rely on
A> file system layout (e.g. /etc/fstab is gone), it must not rely on kernel data
A> structures, and it should be careful with system calls.
A>
A> You could ask the 3.rd party software providers, or ask you Sun salesperson to
A> check them out in Sun's catalyst catalog.
A>
A> I have been running Solaris 2.0 on my workstation for a while now.
A> As a user, I see no difference (as a user, I do all my work within a windowing
A> system, and OpenWindows is unchanged. I run most apps without recompiling them.
A> Only FrameMaker and Contool refuse to work in the binary compatibility
A> package.
B> Many (most?) old binaries will run.
B> All the commercial software will quickly be available as
B> S2.0 binaries anyway.
C> Yup, you can take most sun4 *dynamic* binaries and run them under
C> Solaris. Unless they use system calls that don't exist in System V
C> (e.g., getpagesize()).
D> No. Most Solaris 1.x binaries will run under Solaris 2.x, however
D> some things are very different (sV style system administration, for
D> instance). Anything that thinks it knows about stuff in /etc or that
D> has its own device drivers is probably going to fail.
E> Sun is providing compatibility mode features
F> From what I gather, no, they are *not* claiming backward compatibility.
G> Not complete. Some programms work if you install the backward
G> compatibility package, but these are supposed to be compiled with shared
G> /usr/5lib (sys v) libs on /usr/5bin/cc under sunos 4.1.x.
G>
G> >From the few I've tried, simpler things work. Complex ones don't.
G> gcc doesn't work (surprise surprise)
G> X11R5 doesn't work (It is rather complex).
G>
G> Both the above two were *NOT* compiled with sys v libs, but then neither
G> would most sun applications. (I.e. I didn't expect them to work).
H> No. There is a limited "compatibility mode" which will exist in the
H> early releases, but the only supported option is to port the code to
H> SVR4.
I> Nope. Only about half the Catalyst apps were able to run under
I> Solaris without porting. Solaris has a different object format.
3) Does anyone have any positive or negative experience
with early versions of solaris?
-------------------------------------------------------
A> As a systems administrator, I see quite a few differences. I think the new
A> environment is better. I had to read some manuals to get printer spooling and
A> NIS+
A> going. But most administration after that is NIS+ administration that i do
A> through admintool.
A>
A> As a developer, I don't know yet. I'm installing the gcc C-compiler
A> now. I'll see how different things are.
B> It's quite stable, I hear. It's different from SunOS, being SVR4
B> oriented.
C> I'd encourage you to avoid it. I can count on the Solaris 2.1 kernel
C> panicing on me at least once every three-four days (I have to work on
C> a sparc2 running it because we need one for our ports). I don't do
C> anything weird on it, I just use it very heavily. The NeWS window
C> server keeps the load at 1, because of the overhead in doing redraws
C> and such. GNU Emacs hasn't been ported to it yet. GNU C works (Sun's
C> no longer bundling a C compiler), and C++ will work by November 1.
C> Many things haven't yet been ported to work on it.
D> I assume you mean Solaris 2.0 here. I don't, but I've heard good
D> things, though I think it'll be about a year before most people
D> switch. I know that's the time table around here. Part of the
D> problem for us is that Solbourne won't fully support it until late '93
D> (we've got mostly Solbournes). We'd like to switch sooner, but I'd
D> rather wait until it's stable...
E> We're just beginning...
F> I personally have not, though I have heard from beta testers. Their opinion
F> of it usually is "Buggier than Maine in June". Also, I have heard that what
F> does work well is extremely slow.
F>
F> If it is released soon, I know we (at Kodak) will stick with SunOS until
F> Solaris 2.1 or 2.2 is out.
G> I've installed it on an ELC which I am using to "stuff around" on in my
G> spare time. The first thing I had to do was get a binary copy of gcc
G> (ftp.uu.net:packages/cygnus (or some such directory)). This extracted off
G> and works pretty good. The binary archive is about 5.9 megs, the src
G> is 10megs. The binary includes gcc, gdb, and other usefull stuff.
G> It *does not* contain make (i.e. grab that as well).
G>
G> I am fortunate in that at work we use System V and BSD so the migration is
G> fairly easy. If you only know BSD, some of the commands are subtly different
G> - no more "shutdown -h now" its "shutdown -i0 -g0 -y"
H> It's basically vanilla SVR4.
I> See Open Systems Today review of 9/28 by Jason Levitt.
4) What are the answers to questions I didn't know enough to ask?
-----------------------------------------------------------------
B> Is it going to be a commercial success? I think so.
B>
B> Solaris 2.0 on {SPARC and 386} will be the first time Sun crosses an
B> installed base of a million machines. That will happen roughly in
B> Summer 1993. Most people agree that an installed base of 1e6 is where
B> floodgates of 3rd party work open. We can expect the software
B> packages available to go up by 5x from the current numbers of around
B> 4000.
B>
B> >From the viewpoint of software development, the thing to do right now
B> is buy a SPARC box and start writing software. The moment S2 for 386
B> ships you can recompile for 386.
C> The major difference will be the transition your sysadmins will have
C> to make. The rules completely change (printing's different, terminal
C> management, startup files, NFS, everything).
D> The biggest question is probably the support the vendor will give to
D> Solaris 2.0. When will they support it? How long will they continue
D> support of old stuff?
E> Ask Sun for a copy of their migration kit and for a copy of Solaris 2.0.
E> Solaris 2.0 is intended for customers to try out in a *non-production*
E> environment. Solaris 2.1, the first production version is due in
E> Nov/Dec.
F> Difficult to say, really, without knowing the nature of your organization.
F> Will it consider a vendor other than Sun, HP or IBM? Will it consider free-
F> ware alternatives? If so, does it have the necessary software people to
F> maintain and customize the freeware alternatives?
G> Of course all this is personal opinion, but at work we have Mips M120s,
G> HP apollos, rs6000s, suns and [34]86 boxes. My personal favourite would
G> have to be the suns. I like the OS (4.1). I like some of the things that
G> SysV R4 is planning. Soon we will have it on the Mips' and this will make
G> it easier for stupid users to remember commands across different platforms.
G>
G> >From what I have stuffed around on, I think solaris is pretty cute. Of course,
G> I have just started my playing, and we do development rather than "production"
G> type stuff, so I don't need to stick my neck out be releasing it to critical
G> applications.
G>
G> Of all the Unix's I have played on (quite a few - I port a RDMBS for our
G> company) I like AIX and A/UX the least. This probably stems from my hatred
G> of IBM saying "We are big. This is the way we do it. This should be the
G> Unix standard". And I don't like apples GUI copyright attempts.
H> You will need Solaris 2.x to run the multiprocessor SS10 machines.
H>
H> You will lose a lot of the BSD-derived commands.
H>
H> You also don't get a bundled C compiler, but gcc 2.x binaries will be
H> available and you _DO_ get the Sun system header files and libraries.
H>
H> Ask your Sun rep for a copy of "Pipeline to Solaris 2.0."
Thanks to the contributors:
A> Birger A. Wathne, birger@ii.uib.no
B> Ajay Shah, ajayshah@usc.edu
C> Brendan Kehoe, brendan@cygnus.com
D> Christopher J. Calabrese,cjc@ulysses.att.com
E> Bruce Cogan, Bruce.Cogan@anu.edu.au
F> Brian Talley, talley@acadia.kodak.com
G> Geoff Newton, gjn@cs.uq.oz.au
H> James W. Adams,jwa@alw.nih.gov
I> Andy, andyfe@utoday.com
-- _____________________________________________________________________ R i c k C o t t l e M O T O R O L A, I N C. Project Leader,Factory Info. Systems Final Manufacturing Operations Mail Drop EL606 Logic & Analog Tech.Group Email:RRBK50@email.sps.mot.com 2100 East Elliot Road (602) 897-3867 Fax:(602) 897-3608 Tempe, Arizona 85284
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:50 CDT