Exactly.
> new installs. But they did not, probably because the Win95 and NT teams
> do not work very closely together and in some cases do not even like each
> other. This is going to cost them big time in support costs and user
> complaints, especially down the road, because VFAT does not take well to
> large accumulations of files and is rather fragile compared to "real"
> file systems like NTFS, HPFS, or our own ext2.
Yep. I've already had some corruption start to happen. Got a bunch of
lost clusters last time I ran the scandisk whatchamacallit, and a certain
DLL doesn't work anymore (must have been one of the bad clusters). My
ethernet driver never works the first time Win95 loads but does work if
I restart windows. etc. But it sure is a sexy UI! :-)
>
> There have also been mixed signals on whether NT (i.e. the corporate
> heavy-duty OS) will ever merge with Win9x (i.e. the light-duty/home user
> OS). It may never happen, and even if it does it is not yet in sight.
Yeah I heard Billy said it would happen, then a few days later said it
wouldn't for the foreseeable future.
>
> > Isn't NTFS basically sound?
>
> Yes, but not anything like a superset of ext2. Probably Linux can be
In what ways is it inferior?
> kludged to run on it a little more elegantly than UMSDOS. I think it
> does make a great deal of sense to work hard on read-write NTFS support.
>
> > NT supports HPFS too as I recall.... it could easily enough become the
> > lowest-common denominator for cross-platform FS's. Seems like there was
> > a reason NTFS was better for NT than HPFS though, HPFS didn't support
> > ownership or something like that.
>
> Nope, sorry, NT 4.0 drops HPFS support. Besides, OS/2 will run in FAT and
Bummer. It might just be an act of antagonism too - that would be like them.
> has a rather dubious future, unfortunately, despite being technically
> better than Win95 in many ways. IBM has snatched defeat from the jaws of
Yeah, dammit.
> victory. From the numbers I've seen, OS/2's user base is about the same
> as Win95's and is now falling off; meanwhile NT probably has fewer users
> than Linux but is likely to get a big boost when 4.0 *finally* comes out,
> perhaps this summer.
>
> [Meandering mode on]
>
> In my opinion the only thing really important NT has going for it is its
> ease of use. Performance is not particularly good; applications are sort
End users can actually administrate the machine. Let's face it, that's not
true of any Unices I know of; you have to know an awful lot of "guts" to
get anywhere with Unix. Sometimes I wonder why it was deemed a good idea
to create a free Unix instead of a free "ultimate operating system", designed
from scratch, but maybe backwards compatible in its libararies so that Unix
source could still be compiled for it. Just the old evolution-vs.-revolution
thing I guess; it's a whole lot easier to improve things than re-invent
them, especially when there's a big code base already available for Unix,
and especially if you're learning how as you go. Be on the other hand is
writing theirs from scratch, and if they do it right it could be wonderful
for the hackers and software designers but get orphaned like the NeXT for
lack of users.
> of there but are no more advanced than those of, say, Solaris 2.x; it is
> less scalable than most Un*xes; and its Windows 3.1 support is limited and
> not very desirable to use. It is getting most of its installed base with
> people who don't want to learn Un*x and are used to Windows from the
> desktop. It will be interesting to see if Caldera can make a dent...
Yeah WinNT was another "ultimate OS"; backwards compatibility was low-priority.
They hoped they had the clout to make it a market share winner, but then
sabotaged their own efforts by not making it "the one" OS. We develop
OS/2 software where I work, and the non-OS/2 zealots say that NT has the
robustness of OS/2 without being an oddball and with a better UI; while one
guy in particular who is an OS/2 zealot says the only reason NT is as good
as it is is that you-know-who incorporated a lot of actual OS/2 code into their
product. (He believes the reason the regular Windows series isn't as good as
NT is that y.k.w. doesn't understand their stolen OS/2 codebase well enough to
reproduce the results on their own, only just add a thin veneer of nice-
looking UI stuff over an existing product.) OS/2 is indeed very complete (the
kernel provides more services than Windows) and robust (its scheduler is
possibly the best in any OS, true pre-emptive and all that) and generally a
good platform to develop on. But so huge, and so lacking in support. In a
way you kinda can't blame IBM for that.
Ah Caldera. Well, it's still Unix, and still un-free. Our OS/2 zealot here
also says Unix is too hopelessly old, and if it was going to "catch on" it
would have done so long ago (having pre-dated CP/M, after all). Something
about its multi-process scheduling being based on 1-second timers instead of
OS/2's 32-millisecond timers, thus making Unix useless for real-time stuff.
(As if that was such a major architectural feature we couldn't change it...)
And face it, like I said above, the complicated administration sucks. But
then, OS's which present a friendly UI must by necessity still be just as
complicated in their guts, and then if the guts get broke you can't fix
it; you have to wait for the glacial development giant to release the next
version. Guys like Caldera can make things easier for users by guifying
all the nasty stuff. But X is also too complicated and a same or
worse memory hog as OS/2. And its API isn't even OO! At least in OS/2,
to create an event handler I can just inherit from a library one, and
re-define the few member functions that I actually need to redefine.
As I see it X's only advantage is being the only GUI I know of that has
the built-in client/server concept (compute on the workhorse and display
on the workstation). That's a cool concept. One which CORBA may make
obsolete in X's present incarnation....But it will take a few years for
ORB to be built-in and easy to use with any major OS, I betcha. MS is
still trying to be tricky and proprietary with their own solutions (like OLE)
instead of working with the rest of the world on open platforms. Now
there is going to be so much more consumer demand for client/server tech
though, among the common folks, as the Internet catches on, and Java, and
cheapie terminals with massive computing power but no storage, and intelligent
agents and Telescript and all that, that the demand may be enough to create
a whole lot of competition and we may see in 5 or 10 years what we would
today consider a real OS. I hope so. And then we'll be yearning with all
our hearts after some other new idea. I keep thinking as if specific
technologies will mature and become stable, but I guess they never will.
It frustrates me to no end that I have yet to see an OS I liked whole-
heartedly. Taligent might have been it if they hadn't given up. I think
what I want is a really dumb, fast microkernel written in assembler with
OS/2 like multitasking (smaller timeslices though). Then you build a huge
"rest of the kernel" layer
that consists mostly of DLL's in small, logical chunks so that many of them
don't have to be loaded all the time. Most drivers, the ORB implementation,
TCP stuff, all the GUI/graphics functions, etc. are all in that layer.
That layer even has sublayers for various parts of the same functionality
(to separate video acclerator drivers from graphics primitives from 3d
extensions from GUI elements from event handlers from pre-built dialog
boxes from applets, for example). Everything that ever could be done
is all thought about in advance and provided for, with an empty implementation
if nothing else. It's all 100% OO, 100% extensible and dispensible (if a
DLL is incomplete or buggy it should stay isolated to that DLL). Heck, maybe
an interpreted or tokenized language would even be approriate for this layer,
so that it's more user-editable and user-extensible. All these DLL's are
CORBA-compliant so that you can just grab certain layers and put them on
remote machines. Pick exactly how much processing you're going to do
on your X-term and how much on the server (no, not the client, I know that
is what they call it in X-land) just by how much of the extended kernel
layer you chop off and put on the remote workstation. You don't need a
standards body to define something like X protocol, because it all "just
happens" through the magic of the ORB. _All_ function calls are remoteable
not just the ones to draw graphics primitives. And its filesystem is OFS
(my invention - I'm going to actually implement a first cut at it one of
these days I think) making AI extensions easy to add as the algorithms
get better.
[Meandering mode off - hey man you started it! And left your closing tag
off too. Gimme an inch....] :-)
-- _______ KB7PWD @ KC7Y.AZ.US.NOAM ecloud@goodnet.com (_ | |_) html: http://www.goodnet.com/~ecloud __) | | \__________________________________________________________________ * Interpedia * ARS * electronic music * cyberspace * ham radio * sci fi *