On Fri, 15 Jan 1999, Herbert Wengatz 42850 wrote:
> At that time Sun's intention was that an OS is only the vehicle
> for selling hardware. Not a product itself, since they didn't
> wanted to make themselves concurrence.
I agree that this was their strategy and intention -- it was just a
stupid strategy, and pretty obviously so. It was a very common strategy
(and mistake) though that has been made many times and is still being
made. S. Jobs, for example, took years to figure out that he was
trying to sell software, not hardware, and by the time he learned it his
hardware company was ruined, his software venture was being supported by
previous rivals (e.g. Sun), and the combination of religiously defended
mistakes in NextStep (HopeI'veGotTheCapsRight;-) and clones of its best
features made it too late for a "Next" alternative to Windows. However,
there was a brief moment in the history of systems when NeXTStep, sold
for $50 complete and ready to run for PC's, would have blown Windows and
Microsoft through the wall and into the next county as well.
Apple is still making the mistake, although they are gradually working
their way out of it. Unfortunately, Apple's OS's have traditionally
been crippled by religious fervor (Death before a command line
interface, Infidels!) and the fact that it was (and still is) trying to
evolve into a real multitasking operating system with only one
integrated GUI system interface. Kind of sad, really. Apple's efforts
have a Frankensteinian cast to them, as if a mad scientist tried to
create a human out of spare parts from a generic description of a human,
leaving out all the things deemed by the M.S. as irrelevant, without
ever passing through an evolutionary phase to test the utility of each
"improvement" (or omission). Makes one really appreciate Darwin...
They just don't get it. Hardware has been a commodity for over a decade
now. Consumers (corporate, personal, and other) rarely give a damn
about which CPU they use. They care about price point (can I afford it
regardless of the features), price performance (OK, I can afford
SOMETHING, now how much can I get for my money), availability of "key"
applications (PC's sold like hot cakes back when they could run Lotus
123, Wordstar, and maybe DBase II on top of DOS, for gosh sake), and
LAST ease of use. Only borderline luddites who secretly feared
technology, the wealthy who didn't care about price performance, or
groups with a high fraction of users who were technology impaired (kids,
secretaries, mothers-in-law) ever bought Apples -- everybody else bought
what they could afford (usually a PC clone) and learned to live with
DOS, which was pretty easy because all they had to do was learn to
launch a few applications. Like type in "123<CR>" or "WS<CR>" instead
of clicking on something. Hey, even computer dummies could handle this,
and once inside the applications were all self-teaching.
I'd best be careful, though, or I'll get off on another tirade. The
real point is, times really haven't changed. Sure, various things have
gotten much better. Everybody has a GUI now, Apple's isn't even the
best one anymore. There is a LOT of software, and more choices in
hardware. Still, users want to buy a system ready to run "a bunch of
useful stuff" for as little money as possible while getting the most for
their money, and they hate like poison learning things over and over
again -- they want their learning curve to have some long term advantage
when they get over it, as in maybe survive a few hardware upgrade
cycles.
They haven't gotten this from Windows. 3.x->95->98->200? has proven to
be a pain in the butt for users, as each generation is significantly
different from the preceding one, each generation basically won't run on
the hardware of the previous generation, each generation does a poor job
running software written for previous version, and each generation has
lousy features and a rich set of bugs.
I'm not qualified to talk about Apple's OS (and don't want to be the
target of a jihad:-).
They are, however, in a position to get this from linux, if a certain
degree of caution is exercised in altering the kernel's interface (which
makes this a suitable topic for linux-smp after all). The good news
with linux is that it runs on anything down to a 386 and runs on nearly
every major CPU architecture short of big iron (and if there were enough
big iron systems around where people wanted it, somebody would probably
port it there;-). With a clear functional differentiation between
OS/kernel and both the GUI and application layer, it isn't hamstrung
like NextStep or Apple's various OS's (and now Windows, which has
repeated the errors of its elders by eliminating a command line/shell
interface that is LOWER than the GUI itself). Consequently, when one
learns linux (with an admittedly steep learning curve!) one won't
suddenly find that one's applications or OS interface radically change
when one buys new hardware or upgrade the OS (distribution or kernel).
This brings up a few caveats. Some parts of the kernel interface are
still evolving. For example, going from 2.0.x -> 2.2.x, there are some
fairly significant changes in e.g. /proc. I suppose that this is
inevitable, evolutionary and a good thing, but it is this SORT of thing
that can confuse and annoy a basically conservative owner/user. For
that reason, it would be most wise to really think ahead on this and
work out a SCALABLE and EXTENSIBLE /proc layout that can accomodate new
information without screwing up or radically changing /proc-parsing
applications. A bit of a challenge, but clearly 2.0->2.2 broke a lot of
/proc parsing stuff and it would be good not to do this every 2-3 years.
The same sort of thing also holds for libraries. As a long time Unix
user, one of my pet peeves is a perfectly good application that does
something really useful that won't recompile without major surgery after
somebody decides to completely revamp one or more of its core libraries.
Don't get me wrong -- in many cases, the changes are for good reasons
(like moving toward posix compliance or to accomodate new improvements
in the kernel or hardware) but they cost TIME and hence MONEY or
ANNOYANCE to repair. I think that many of us have encountered this
recently as we have worked to swallow the libc 5 to libc 6 (glibc)
transition. Stuff breaks. Unix experts can cope, but Joe Consumer
won't be able to, and even the protective layer provided by
"distributions" won't help Joe understand why application X won't run on
his system.
Having recently become a genuine Red Hat Linux user (after years as a
Slackware person), I can now state somewhat authoritatively that RPM's
are not the answer to this. They may help, sure, but even with packaged
CD sets and scripted installs, it is way too easy to get a dreaded
"needs library whatever" message when installing an RPM and be left
scratching one's head as one tries to figure out where the devil one can
get library whatever if it didn't come with the distribution (or
prebuilt in the RPM) in the first place.
Almost all the ingredients are there for linux to accomplish its long
awaited world domination. Tools like linuxconf are pretty good (as good
as anything available for Windows, and that is good enough for now).
The management of daemons still sucks -- if the daemon isn't
preconfigured and preinstalled for you, it is going to be a painful
process at best even for benign and common ones like printer daemons.
One has a fine selection of applications available right "out of the
box". It would still be lovely to have boxed games for linux sold in
over the counter software stores (when this occurs, linux will
officially have "arrived"). I do hope that effort continues to be
invested in creating an idiot-proof layer between the OS and the novice
user/administrator. Distributions like Red Hat have this layer as their
primary "advantage", but the price one pays for this is that it is
virtually impossible for non-expert systems persons to work outside the
tools they provide, and if those tools are inadequate (and they are
still inadequate in many areas, I'm sorry to say) you will have
frustrated users giving up and going back to Evil Empire operating
systems.
> But I still hang on my dreams how great the world could be,
> if everybody would have an Workstation with an old SunOS...
Yeah, for a while that was a good dream...;-)
> The two reasons for why I don't use FreeBSD instead of Linux
> are:
> 1.) Linux seems to be best of both worlds, SysV and
> BSD. And since I'm working (at work) with SysV,
> I don't want to have a change in behaviour between
> home and office.
> 2.) The drivers for my PC-hardware are much earlier
> available. If I do buy me a brand new XXX-featured
> graphic-card or SCSI-controller, or whatever,
> I don't want to wait for two years until the noble
> "house of lords" have finished their discussions
> and start implementing a driver.
> At that time my "brand new" hardware is from the
> stoneages and I have a new board in my hands, which
> is currently not supported. - Got the hang of it? ;)
>
> However, those two points are crucial for ME. So I don't want to
> break any religiuos wars loose here. At least, we're in a Linux-
> list, aren't we ? ;-)
I think that this is why a LOT of linux users prefer linux. linux is
less controlled and more chaotic than freebsd. More choices, less
stable (perhaps:-) faster evolution. I personally think faster
evolution is better, as long as linux doesn't develop cancer, which is
unlikely as long as we rigorously weed out the unfit ideas and
keep/exploit the fit ones. The analogy, by the way, isn't all that
strained -- the OS wars are a big, real-life experiment in genetic
optimization that would have made Adam Smith and his Invisible Hand very
happy...
rgb
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:[EMAIL PROTECTED]
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
