Date: Wed, 8 Oct 1997 23:23:13 -0600 (MDT) From: Wes Peters <softweyr@xmission.com> To: "Jordan K. Hubbard" <jkh@time.cdrom.com> Cc: chat@freebsd.org Subject: Re: Digital, Intel, Silicon Graphics (fwd) Message-ID: <199710090523.XAA01838@obie.softweyr.ml.org> In-Reply-To: <26258.876324035@time.cdrom.com> References: <199710081200.IAA17975@hda.hda.com> <26258.876324035@time.cdrom.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Jordan K. Hubbard writes:
> Unix also lost any chance it might have had for "total victory" about
> 10 years ago, when it Balkanized itself instead of fighting its
> external threats, so why even worry about that particular lost cause?
> I still think that we can find some measure of success in turning away
> from the main axis of the battle and looking for smaller pockets of
> territory to capture, things which require much higher-tech solutions
> than Redmond is able to provide.
Unfortunately, the trade press in general still sees "UNIX losing the
desktop war to Microsoft" as a bad thing. I suggest the opposite is
true: UNIX was never meant for a word processor/spreadsheet system.
Since we never really wanted to go there, why would we lament somebody
else inheriting that stick ball of goo?
It's pretty easy to argue that Win95 wasn't the "best" contender,
compared to (insert your favorite "desktop" OS here), but UNIX wasn't
even in the contest.
Let's get on with making UNIX absolutely killer at what it was designed
for: being the best multitasking, multiuser, multiwhatever you want
small computer system ever designed by, for, and of programmers.
> I think that technologies like vxWorks and QNX are doing pretty well
> for themselves in the embedded systems market, for example, and much
Oh yeah. Of course, Microsoft has now decided to take over this market
too. The Embedded Systems Conference last week in San Jose had "Windows
CE" plastered all over it. Fortunately, most of the people there were
pretty skeptical, and nobody appeared to be taking CE seriously for
anything other than handheld web browsers for managing embedded
systems. This, of course, speaks well for those of us in the business
of making embedded web technology. ;^)
As a comparison, Wind River was loudly and proudly telling everyone at
the show that they own *100%* of the operating system market on Mars.
;^)
> of the core technology in Unix could be highly applicable to this same
> sort of work if some serious attention were paid to structuring it
> more as a set of pluggable components and providing better real-time
> capabilities for the people who need that sort of thing. FreeBSD in
> the recording studio, anyone? :) Seriously - such markets may be
> comparatively small but they're still enough to keep a fair number of
> Unix hackers gainfully employed.
Comparitively small? Heh. With WRS shipping on the order of a million
licenses a month? Of course, they're something like 45% of their
market, but the numbers are pretty impressive. Think about the sheer
numbers of things that have embedded OSs in them these days: laser
printers, microwave ovens, VCRs, cell phones, cars (typically 4 to 9
systems per car!), medical instruments, etc. And, of course, network
routers, switches, print servers, etc. My previous employer made
studio-quality A/V switchers for TV stations, using VxWorks for the
control systems and a combination of HP-UX and *FreeBSD* for software
development.
> If you asked me what I thought the real challenge ahead for Unix was,
> I'd say it was in fulfilling its own original promise, not in trying
> to become Billy's personal nightmare. We need to take careful stock
> of the various shortcomings which are standing in the way of Unix's
> becoming the ultimate engineer's toolbox and solve them. We need to
Well spoken. We don't need UNIX to be everything to everyone, all we
need it to be is what programmers *want* to use. If I can develop it
quicker, faster, easier, more maintainable on UNIX, and UNIX runs on
whatever hardware I'm likely to get my hands on, I'll pick UNIX. If
your "boss" doesn't allow you that leeway, wants to put roadblocks in
your way, VOTE WITH YOUR FEET!
> take the best ideas from the embedded OSes, like having
> dynamically-loadable-everything, a light-weight GUI subsystem, POSIX
> real-time extensions, etc. and implement them. We need to fix Unix's
Along the way, I think you'll ultimately improve the system for
"general" users as well. Plug and play devices? Pretty simple, if all
of your device drivers are dynamically loadable *and* unloadable from
the start. Multithreaded servers for speed? Pretty simple if the
kernel itself is multi-threaded, you just expose the threads API to user
code as well.
> existing services so that they Just Work and you don't have big
> warning signs over things like remote file locking which say "out of
> order" and cause an engineer to mistrust his tools. Make the
> components easily separable so that it's possible to have only as much
> "Unix" as you want for a given application. Make it all fast as heck.
> Play to Unix's existing strengths, basically, and continue to innovate
> along the same lines that our spiritual forefathers (so to speak) had
> in mind. I think the original folks were after a really cool toolbox,
> basically, and they succeeded admirably enough to draw a whole host of
> other tool-users into what they hardly expected would quickly become a
> Craftsmen's Cult of sorts. ;-)
One of the hardest things to convince Windows programmers of in moving
*to* UNIX is the tool-building cult. They come from a world where you
buy an all-encompassing, all-singing, all-dancing single IDE, and if it
doesn't do it, it can't be done. And, of course, they pay a lot of
dollars for these monstrosities.
UNIX, on the other hand, offers lots of tools, small and large, and a
pretty straightforward way to tie them together: scripting languages.
I use a configuration management tool called Perforce (highly
recommended, by the way) which runs on Windows and a variety of UNIX
systems. The servers are available for both UNIX and NT. It's pretty
amusing to watch a user ask a simple "how can I automate this task?"
question, see several more experienced users contribute 5 and 10 line
shell scripts that accomplish the given task, and then 2 or 3
unfortunate souls say "we use the Win95 client, how can I do this in the
GUI?"
If I weren't cruel and capricious, I could write a Tcl/Tk script for
them. On the other hand, if they had any smarts, they'd throw a bunch
of $$$ at me so I'd feel better about doing it. ;^)
Another marketing channel we FreeBSDers must concentrate on is the
university. UNIX originally climbed into commercial existence through
the influence of young programmers entering the work force with UNIX
experience from their colleges and universities. These days, colleges
buy the "educational discount" versions of VC++ and teach their
graduates nothing but "how to write an MFC app." Talk to professors and
students at your {current, former} universities. Get them to use
FreeBSD. Get them to install FreeBSD ftp servers on the campus net.
Recruit a Doug White at each and every college. Get them to stock
FreeBSD CD-ROMs and books in the bookstore. And get them to take
FreeBSD to work with them. ;^)
> If we're going to carve out and hold a credible niche for ourselves in
> Mr. Bill's world, we're going need to get back to the fundamentals and
> stop thinking about painting a happy face on the outside of the
> toolbox as much as making the tools inside high quality instruments of
> software craftsmanship. That's where "commercial UNIX" essentially
> went down the wrong path, in my opinion. Marketing was brought in
> without any clear idea as to what it was they were trying to sell and
> so they kind of looked over their shoulders and tried to sell it like
> the other guys were selling their OSes. Unfortunately, the "other
> guys" were Microsoft, IBM and Apple in this case and very bad examples
> for our impressionable little marketdroids, their requests eventually
> resulting in the diversion of engineering resources into a battle for
> the desktop which could not be won and should never have been fought
> in the first place. They went astray and they paid the price.
>
> Anyway, maybe that's our secret weapon. No marketing department. ;-)
Exactly so. Every time I've seen one of my products screwed into the
ground, it's been at the hands of "Must have feature implement by Q1,
10,000 license sale pending." Then you give them the feature they
wanted, and they never bother to use it. Gah!
--
"Where am I, and what am I doing in this handbasket?"
Wes Peters Softweyr LLC
http://www.xmission.com/~softweyr softweyr@xmission.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199710090523.XAA01838>