#CBM layoffs
30 messages in this thread
If the UNIX is the only OS you need to run the A3000UX is definatly NOT the
machine to buy!! Unix SVR4 is supported by at least four different vendors
on the Intel 386/346 platform I know a multiuser SVR4 development
environmeny can be purchased for under $2000 from UHC or ESIX. Also is a
rubust selection of software available for SVR4-368/486 where for
SVR4-68030 there is next to none.
Yeah, but you'd still have to run on an Intel processor. Gross. I'd
rather debug my drivers on a 68K or 88K machine anyday.
Steve
If the only OS I need to run is UNIX I don't care wich processor it runs on
just so its quick and there is software available. While the UNIX may not
run as fast on a 386/486 as say a SPARC or RS6000, it is a fraction of the
price and just as much (if not more) software available.
It really depends on what you want to do or run…..I have access to
hundreds of megabytes of PD software for UNIX (in source) for everything
I'm interested in….so it doesn't matter what processor it uses. Since I
write UNIX software, and work on things in the OS, I'd rather use the
Motorola processors.
Besides with machines having ABI (for 68K) and BCS (for 88K) compliance,
there are many packages available. And you're not limited to a particular
UNIX implementation.
Steve
David,
I need a machine to run a small wholesale distribution company. The
software system runs on many machines including 386/486 and the A3000UX.
The A3000UX doesn't seem too expansive when compared to a quality 486
machine configured similarly (300meg HD, 8mb RAM, 150mb tape, ~16 serial
ports). I will initially have 10 terminals. This will not be a
development system. Most likely it will end up being a 486 machine anyway.
Sounds like you are familiar with such machines. I have not heard of
UHC or ESIX. I have heard of SCO. The availability of other programs to
run on the machine is not needed at first but will probably be needed on
down the road.
….Mark
The main problem with the A3000UX is not hardware but getting the software
vendors to bother take the trouble to recompile and test their software for
the five or ten people with Amiga Unix machines. The software you need now
may be available for the A3000UX, but what about what the software you may
need in the future? You can count on the fact that the platform with the
least number of users will recieve the least amount of attention from
software and hardware vendors.
You've hit on exactly why they've come up with the Application Binary
Interface. Write software that follows the ABI standard, and it runs on
all ABI compliant machines. You're not stuck if you just have SCO UNIX
(for example) and the software you want to run hasn't been ported.
At one point, a binary from a Motorola machine (Delta maybe?) was brought
over to the Amiga and it ran just fine.
Unfortunately, because they've found a problem with the ABI implementation
on Amiga UNIX, that will be (hopefully) fixed in a future release
(supposedly 2.1 of UNIX). (Figures, doesn't it?) Probably runs most ABI
compliant software, though.
Steve
Steve
ABI and all that sounds nice, but are there any other ABI compliant
implementations of SVR4 for the 68000? I don't know of any. If thats the
case as far as commercial software is concerned ABI doesn't give you squat!
As far as loving the 68000 processor, Well I have done programming on unix
machines with 386, 68030, and Sparc processors and have found that with
unix the processor makes little difference. Thats the whole idea of unix,
to be hardware independent.
David,
And you've just given a very good reason not to worry about
programs written specifically for 68k based machines. ABI compliant code
for a 386/486 machine should run on a 68k machine with a simple re-compile.
In theory…no port (other than the recompile) would need to be
done.
Don
Check that again, Don… (or am I missing something `:-\
ABI (I forget what that stands for) is a binary standard. That means that
a program conforming to the standard will be an executable that can run on
any similar hardware. In other words, if some vendor has an executable for
(say) a Sun-3, it should work on an A3000 (maybe all Amigas) with *no*
recompile. In fact, even if you had the Sun-3 source code, the ABI doesn't
guarantee that it will compile on an ABI compliant machine. The key with
the ABI is that the machine hardware (CPUs, I think) are the same (so ABI
compliant 386/486 executable is not expected to work on a 68K).
The other technique that OSF is backing is ANDF. In this case, vendors
compile their code into an intermediate (and unintelligible) form by an
ANDF compatible compiler and that is then distributed that to the users.
The users then run this intermediate form through a final compiler that is
specific to their machine and (voila) they have their executable. The
theory is that vendors need only develop their code once and can instantly
distribute to all ANDF compliant machines.
The question mark in both cases is how well the result will be able to take
advantage of any special features of the end-system (which is the bread and
butter of the Amiga).
David
– via Whap!
David,
Yes and no. ABI stands for Application Binary Interface. Thus, as
you say, an ABI compliant program compiled on a 68k machine should run on
any other ABI compliant 68k machine as is. However, part of the interface
spec is that is uses "generic" system calls (calls that run in kernel mode
and "hit the hardware") and the kernel code takes care of all the specifics
of the hardware. Thus code is not written for a specific CPU, only
compiled on a specific CPU. The code should also run on any other ABI
compliant machine (with a different CPU) by simply re-compiling the code on
a machine with the new type of CPU.
The source code should be totally hardware independent as all the
niggling details of hardware differences should be taken care of in the
kernel interface.
As a very simple example, that already is application independent,
an open() call looks the same on a 486 machine as it does on a 68k machine.
The code in the kernel that does the actual open of a file or device would
be different though.
Don
Don,
Your message implies that the ABI spec also will 'handle' certain
processor differences, like the byte-ordering of multi-byte data elements
'transparently' to the application. Is this true?
That's not been my impression of what ABI was intended to address, but I'm
not particularly involved in the Unix world, so I don't know.
…BobR
Bob,
My understanding is yes, it'll handle things like that…however
it'd be handled by the compiler not the interface. However, I'm sure that
there are ways of writing code so the compiler would get confused…and
some changes would have to be made to get the code to run on a different
processor.
Like all things in life, compilers, interfaces, and the like can
only do some much. The author has to take some responsibility to account
for things like that using #ifdef's and macros if they're really going to
get tricky in things like bit manuiplation.
That's not "authorative", that's just my understaning of how it
works. I've never actually tested it yet.
Don
Well, I haven't read the ABI, so I'm certainly no expert.
This would suggest that the ABI implies a single operating system so that
all the kernal calls will look the same. This makes sense since the ABI is
backed by AT&T and they would be pushing for UNIX as that one operating
system.
I thought that the ABI was just a specification that attempted to ensure
that executable code looked consistent when it got to the CPU as an
executable, but stopped short of how the executable code was translated
from source code. For instance, UNIX might call it an open() in source,
but AmigaDOS might call it Open() in source. The ABI wouldn't care as long
as it looked the same in the final executable.
David
– via Whap!
David,
ABI is a UNIX SVR4 feature, it's got nothing whatsoever to do with
AmigaDOS.
So yes, it's dependant on your machine running SVR4. And of
course, for the binary to work, it's also got to be compiled for your
"family" of CPU.
For example, let's presume Apple ported SVR4 onto the Mac (they
haven't…it' just for sake of example I'm using it). Then you could take
any program (other than unix itself…cause that's wherethe "magic" is
done) and copy it onto an A3000UX (running SVR4…not AmigaDOS) and it'll
run as is. You could not though, take the program file and run it on an
Intel machine even if that machine is also running SVR4.
You *should* however, be able to take your source code for that
program, copy it onto the Intel machine, recompile it…and it should then
run.
Don
Don
I'll idmit that when I first heard about C= comming out with SVR4, I
thought "Now I'll be able to run real Unix apps because of ABI, re-compile
and all that". This is valuable if you can get by with only public domain
software or packages where source is included. This is not the case for
me. The apps that I develop must run with an SQL server. I do not know of
any vendors that have made theirs avalable for the Amiga or 68030 ABI SVR4.
I have checked with SYBASE, ORACLE, INFORMIX, and INGRES and so far no
DICE. I have heard strong rummors about ORACLE comming to ABI 68030, but
only rummors. The all say that they not only have to recompile but train
their support staff as well.
I have heard that UNISYS and perhaps Data General are working on their own
68030 ABI machines. I hope sg. This would definatly wake-up the software
vendors to port their stuff. But if things stay the way they are Amiga
Unix will not be the platform of choice for a serious production system.
David,
My "sources" in Oracle Corp say that they definately are coming out
with an ABI compliant 68030 SVR4 package. No time scale yet. I've also
heard that Unisys and DG are working on SVR4. Again…nothing definate,
and Unisys is in a bit of trouble…no idea on DG.
Yes…you have to train the staff, but you've got to train the
staff for any new OS, and the training for a '486 SVR4 should also carry
over to the 68030 machines since from the user's perspective…they aren't
different machines.
Basic tuning of the machine should be the same…with little, if
any, differneces machine to machine. Oracle itself would run/work the same
on both platforms as it does now across various platforms and OS's (with
some minor exceptions).
But yes…for this instant in time…I personally know of no SQL
based software available for any SVR4 68030 box.
Don
Let's hope fore the best. Unisys and DG joining C= will make SVR4 a
success on 68030.
Just because *you* don't know of any ABI compliant machines doesn't mean
there aren't any. 🙂 I can't name the ABI companies off the top of my
head (besides Motorola and Commodore), but I know that the members of the
88Open group (for the 88K) have many software packages available. (I know
more about the 88Open since I work on the 88K, and I haven't been keeping
up on the ABI stuff).
For applications programming, you're right….the processor makes little
difference…except for the whole Big Endian/Little Endian deal. For
driver development, and OS development, the processor makes a big
difference. Like I said before, programming on the 68K or the 88K is much
better than programming on the 386.
Since that's what I do for a living, that's what I'm interested in working
with.
And as you said, UNIX software is hardware independant, so it's simple for
those companies with "386 UNIX software" to make it "68K UNIX software".
Steve
David,
While I realize your 'five or ten people with Amiga Unix machines'
was an intentional understatement (or I hope it was). I presume you are
aware that the VPI requires it's students to own a 3000UX.
Don
Making software for students with who are up to their head in student
loans? Hardly sounds like a profit making venture <smile>!
David,
I wasn't suggesting that those students would cause publishers to
come out with UX software packages…I was suggesting that there were a
fair number of UX machines out there (on a relative basis), that the UX was
a good enough machine to impress the staff at VPI so they required their
students to buy one, and that there was at least enough software available
that the students could do a bit more than simply boot the machines and say
"wow…it boots up! Now what?"
Don
Actually, Don, What better Unix system to make your students buy? If all
you can do is boot the system, then over the next few years of your course
study you'd HAVE to learn to program unix systems <g>
– Whapping from the North Coast Michael
Dave,
You are right. My biggest concern is support for the software/machine
combination.
Small user base = being ignored. Isn't there some sort of emerging
standard in the UNIX arena for compatability? Well behaved programs would
simply need a compile on the target machine. No doubt I am dreaming.
….Mark
What bothers me is that all you really need is a recompile to move
most all SVR4 software between platforms, and yet it isn't being done.
Commodore doesn't seem able to convince companies to actually
port/recompile.
I saw a nice lengthy list of all the software that is supposedly
ported. About 50 programs, most of which I'd never heard of before.
Admittedly a lot were in subject areas such as accounting with which I'm
not familiar, but there were no big names besides maybe Maple and Comeau
C++.
— Ethan
Ethan:
What accounting software has been ported to the Amiga's UNIX? This is an
area that I'm interested in, as I'd love to get rid of the clones I have at
the office to do the accounting!
-Mike Schiller
In sunny Fort Lauderdale
Mike,
Try contacting: SouthWare Innovations
555 Stage Rd.
P.O. Box 2797
Auburn, Al 36831-2797
205-821-1108
They make integrated business packages for many platforms. The A300UX is
definitely on their compatibility list. I am currently looking at a
complete package for my business and this one looks good. It is not
exactly cheap.
….Mark
I'll dig up that list and post it all, or maybe upload it. Remind
me if I forget.
— Ethan
MAS-90 Series II is a fine accounting package available for DOS and UNIX
machines. Modules include General Ledger, A/P, A/R, Payroll and Time and
Billing. The networked DOS version of the program is my in-house general
ledger, and I'm quite happy with it. The modules are expensive; I believe
they go for $995. each in their UNIX flavor. While I have no information
that they are available for C= UNIX, at that price, they should be able to
port over the code for you, or work out some kind of deal otherwise. I
think if you called up the company (sorry, I forgot their name; I can come
up with that and a phone number if you are interested) and told them you
wanted to drop $3,000 to $5,000 on Amiga UNIX software, I think you would
get their attention. In fact, I think you'd get many companies' attention
if you called them up waving dollars around…
Over the flight pattern… – via Whap!
Ethan,
I wonder what it would take to get a software house to do a recompile?
Can't a dealer for the software just take the source and recompile it using
the compiler for the target machine? The dealer I am working with has the
source as they do modifications for many of the installations they do.
Maybe this isn't true for some of the "BIG" names like Lotus but it should
be for many packages.
….Mark