Ever use a DecPro? Built like tanks, wysiwyg editor with integratable charts and graphics, based on a micro-PDP processor (16-bit flat address space in 1980). Then, some nimrod decided to keep the user locked in this DEC-marketroid approved environment, and that they shouldn't be able to format their own floppies, because DEC would make more money selling pre-formatted ones. They could have just run RSX-11 or RT-11 on it, used the PDP compilers, and instantly had a large installed base with developers everywhere. Instead we got a hacked-together 8-bit processor running a copy of CP/M, and were stuck with that architecture for the next decade and change.
With some vision, they could have been the dominant PC player and become the standard, as they already had a built-in upgrade path, and a decent installed software base. PDP-11 -> VAX -> Alpha Instead, they listened to Ken, marketed wierd machines (still built like tanks) too late (DEC Rainbows), then tried to become a PC company.
What about Wordstar? Glorious program in the early DOS era before WordPervert evolved into a usable product, ran on DOS or CP/M, excellent formatting controls, and didn't need keyboards with either function keys or arrow keys. You could have run it off an ADM3a.
My college, after trying many arguably superior programs (XYWrite, Final Word, Wordstar), mysteriously settled on WordPerfect, probably because a manager thought something driven by func keys that came with a little keyboard template to remind you of what they did was easier. I never understood how a wordprocessor that constantly required you to removed your hands from the home-row keys was supposed to enhance my writing. Then I discovered that nobody can touch-type anyway, so it wasn't making as much difference as I thought.
I'd still probably rather use emacs and TeX, but that wasn't even a dream on early 80s PCs.
Yes, I have, and I still miss them. Your problem was that you (apparently) tried to use Unix on them, rather than VMS, and the common language interface (which allowed you to do system calls and fancy string handling in fortran 77). Once you grokked the Orange Wall (and later Grey Wall), VMS was easy to manage, and rock solid. It used funky networking of course (CMUTEK tcp/ip still gives me shudders), but if you had all VAXes, then DECNET was no big deal. Truly a loss, and superior to many of its successors.
I miss my VAXstation and the 11/785.
turbo pascal 2 was also great, but they never cleanly made the transition to the Windows world. I'm sorry to lose the simplicity of TP2 (which would be great now because you'd just link it to other libraries, rather than rely upon Borland's oddball implementations), and there was always the attempt to be different, such as Turbo Prolog.
Work out at some point what NASA actually spends on science vs the shuttle, and then the alternatives (farm programs, weapons to fight an enemy that no longer exists, office of faith-based initiatives), and you'll see they're not much better off than your field. Coming from an expensive corner of the sciences (chemistry), I used to resent the particle people. 100 authors on a paper, 10s of billions of dollars (and yes, the big ring at Fermilab is way cool, but still...) in construction costs, and a final yield that works out to 1/12th of a fundamental particle/author, if they really saw it. I'm kind of a fan of big telescopes and NASA, as it's the closest most average people come to seeing a frontier and the possibilities therein, and therefore allowing the rest of us to get funded at all. If most people had their way, we'd fund nothing but cancer, fat, and anti-aging research (ok, cynicism off)
Now I'm older and slower, and just sigh and say, "well, at least it's not going to subsidize someone growing rice in a desert".
Propulsion has been solved back in the 60s. The political will to fund the mission and tell the anti-nuke, "let's all go live in unheated mud huts to save the earth" crowd to bugger off is what's lacking. After all, why fund exploration and aspirations when there are farm subsidies and endless petty wars right at hand?
Computers aren't the issue here, mechanical engineering and guts are.
You can get rid of an awful lot of them, though. Despite being stereo only, no home theatre inputs, no remote, my favorite gadget sitting around here is the Harmon-Kardon amp and receiver. Some buttons, and nice big knobs for making adjustments. Tasteful, minimalist readouts. The antithesis of the remote or even front panel on my home theatre receiver.
Basically, your complaints (and I remember the mysterious "box doing nothing with a load of 25" issue" (probably X related), and the replies (recompile everything from scratch after patching), point towards why it's time for you to try *BSD or more likely, Solaris 10. If your boss will spring for new hardware, of course, you should just get a copy of OSX Server with unlimited client access. 15 years ago in Grad school my boss watched me IPL a VMS box from tape, read the console for a few minute, and then emphatically told me that machines like that were supposed to be extinct by now. One might argue the same for OS's which still require you to manually configure the source, and then start recompiling the kernel and libraries. I'm running my lab off an OSX server box, with all the standard services (it manages both the desktops and the compute nodes on the private network), and really would be hard pressed to give a reason that I thought I had to change to Linux, Windows, or anything else. Nobody has ever suggested I recompile anything on that box to make it work right, and there's very little you can't tweak from the gui or through editing rc.local (good old BSD one file for tuning).
A few years back, when Sun was pushing its Linux desktop, it made its sales reps (at least the ones in the upper midwest), use it on their business laptops. Within the year, every one of those laptops had somehow mutated into PowerBooks running OSX, with far happier reps using them. I somehow transitioned from a dual xeon to an old G4 without noticing, because of the apps and seamless experience which just let me work, which is what I suspect you've seen with your Powerbook as well. Someome took the time to add some polish, and it shows.
No, Linux is a hobby. If you ran a real unix (Solaris, AIX), you'd find that same or better stability than you've ever seen with Windows. You'd find better stability yet with Windows if they'd kept Cutler's VMS derived architecture, and didn't shoe-horn home-user Windows features on top of it for compatibility and convenience.
That's true, because if Clippy was true AI, he'd learn from you and become more helpful. OTOH, he'd also have instructions from MS hard-coded in for features that are detrimental to you, but good for their stock price. The end result would be self-confident schizo app (c.f. HAL 9000), who would probably do Bad Things to your documents at an inopportune time.
At 256 M, you're walking (or dragging) XP, not running it. That being said, I've used it on a 400MHz/228 M laptop, and once it's running, it's not too bad. Not desirable, but not bad.
Not to mention the inevitable driver issues. The bad one we hit was certain Trident cards, all claiming to be the same chip, weren't. Minor revisions resulted in different appearances (or working at all) on nominally identical cards bought as a batch. My suggestion to buy ATI or Tseng-based cards only didn't go over well, as it wasn't the cheapest solution.
Yes, but they made such a prettier chime when you rebooted than the Windows box.
Another vote for OS/2. I ran 2.1, then the v3.0 betas through the 4 series. Wrote my thesis using DeScribe on a 386/40. I see you can get Firefox for it; maybe I'll put it on an old Thinkpad for nostalgia, though I'm not certain about wireless support.
Like BeOS, another step into the future, doomed by poor marketing and fighting a pre-installed based of DOS/Windows.
Dragging to the Applications folder was too hard? Sounds like the old Far Side cartoon with two polar bears, an eskimo running off into the distance, and an igloo they just lifted up: "I lift, you grab. Was that just a little too complicated for you Carl?"
I also presume in this case Unintuitive = "Not Gnome".
YMMV, but I ran unices since the 80s, and have gone through everything from TWM (and VWS), through SGI, CDE, and 7 years of Linux on the desktop, including Gnome from 0.3. OSX/Aqua just works. clean, effecient, and doesn't get in the way. Give it another chance, eventually, you'll be assimilated as well, but it's a gentle, California-type assimilation.
I have a Thinkpad A22m from 2000 which is still in active use, so they don't become doorstops after only a few years, if you're sensible. I wouldn't try 3d gaming with it, but given the 1040x1440 screen and a gig of memory, it's more than fine for Word, Photoshop, or web. Laptops are: compact, efficient, quiet, and don't have multiple cables running everywhere. Everything is self-contained, and you can move them if you want to work somewhere else. Now that they come with 17" widescreens for a reasonable price, for most people, there is no reason for an old-style tower or brick design.
Outside of adding another HD or more memory (and even then people take the machine to the shop), nobody outside of geeking circles every opens their desktop system. Given a similar price point, most home users would opt for the space saving design, and simple forklift upgrade when the time came.
NeXTStep isn't dead, it just evolved into OS-X (which given current Mac sales may make it the most prevalent Unix out there). OTOH, you're right about uSoft setting the industry back. In 1987 I could have bought a 386/16 which matched in performance our VAX 11/750 (if you were willing to drop $800 on the math coprocesser). Sun built a Unix workstation around it. Microsoft's answer? 8-bit DOS 5.x. It wasn't until Grad school, when I convinced my boss to buy the Watcom compilers for the lab that we actually made those machines sing. The DOS extender for the Windows 3.0/DOS hold-outs, and native OS/2 32-bit apps, ported straight off the VAX, for everything else.
Then there's the lack of queueing (VMS, late 70s) multiple jobs in the background, and the database-based WinFS that was supposed to ship with Vista (IBM AS/400), Virtualization (IBM, 1960s), TCP/IP (stolen at the last minute from BSD, and relabeled Microsoft TCP/IP), included compiler to encourage tinkering (Unix 1970s, MSDOS 1981). They just have a real bad case of NIH (not invented here), which they should get over.
A former boss who'd dealt professionally with microsoft in the 80s put it, "the problem is, Bill has a vision..... One vision, and it's mired in the 1970s."
btw, the only problem with NeXT is that they were freaking expensive. Mathematica alone must have been $500 of the base price.
Your story about Jim Cordy is similar to the one that was floating around about 10 years back when uSoft put out a unix compatibility layer for NT, and used Korn shell. During the public announcement someone from teh audience pointed out they'd picked one of the more deficient implementations, to which the microsoftie replied in effect, "we believe this version to be just fine". It was then pointed out the critic was David Korn.
Some days I miss the old days, if only because I came from the VAX/VMS side of the fence, and had languages such as VAX Pascal, and VAX FORTRAN, which could be used for systems programming if you wanted (the orange wall had some great examples in Fortran) and were actually readable. I realize that C succeeded in part because it was cheap to free, and came with all Unixes, but it still looks like line-noise to me, and it's still too easy to shoot yourself in the foot with it, not to mention it's barely higher level than some older mini/mainframe assembly languages. Now, because of lazy CSci programs, its successor, C++, is spreading in chemical programming, so I have to deal with mixed C++/Fortran programs. There should just be a rule; if you're doing science, and not writing an OS or games, just use Fortran95. Clean, modern, easy to optimize, talks to libraries such as LAPACK, and relatively safe.
OTOH, I don't miss compiling programs on a 0.6 VUP (~MIPS) shared with 20 other people.
The "not sure personally"... statement gives me pause.
Look, I like (to a point, which is when RMS starts screaming it's GNU/Linux) Linux, it's great for certain applications, and fun to tinker with. However, having just been subjected to two months of intensive Health Care (hospitalized, multiple IV), and spent several of those weeks in a major research hospital (high-tech to the gills, MD/PhD's poking me daily), I really don't want to see something like that deployed until each critical app has been shown to be robust, to communicate well with everything else, and to Not Go Down. My hospital ran my meds and vitals (which are more important than sleep when you're sick, as they'll wake you out of deep REM at 4:00 a.m. to get a blood pressure reading) over a secured wireless system so the nurses had instant access to all the data they needed. They also copied everything on paper and put it in a big, thick, book, which was copied onto the secure optical network. Big instruments (x-ray, etc), had their own embedded, hardened OS.
The issue here is robustness, and it doesn't matter if its SuSE, Win2K server, AIX, or CP/M-86. There is no room whatsoever for political advocacy of an operating environment in a hospital. I'm glad he likes SuSE; I used to use it until it got flaky in the 9.x series, but the OS isn't the issue, the apps are. Show me the apps are stable, and we'll talk about OS advocacy.
We're talking Renaissance operating systems here, so he'll be wearing tall leather boot, a cuirass and morion, with a knife in one hand, and a crossbow in the other. A daVinci-era machine gun is low-efficiency, and requires a horse cart.
Since we can't actually travel to any suspected blackhole/wormhole-entrance, we need the dudes at CERN to whip one up, then someone can step through and hopefully wind up farther away in our own universe.
Linux has mindshare, and by supporting Linux on lower-end systems, IBM sells businesses Power-based small systems (only available from them), and then gently moves them up the food chain as their needs increase. This way you have IBM 2-core OpenPower systems, and IBM p595s, and eventually (salesmen look off into the distance picturing the bahamas) a Z-series mainframe, all capable of being partitioned to still run the same apps. It's part of their policy of making sure they have a solution for any size of business, or any size problem a business has.
AIX, possibly deservedly, has a reputation as one of the 'odd' Unices. Embracing Linux allowed IBM to appear more interoperable and less threatening to potential customers who remember the days of IBM and the BUNCH.
Actually, I run almost nothing but at this point. OS 10.4 on G5s (I know, the graphical front end is 32-bit, but my apps in the background are mainly 64), and I still have time on a pair of IA-64 systems. Remember, Power is not PPC. PPC is a cut-down version for low-end systems, and IBM was probably happy to get rid of a low-volume, high-demand, customer. Real Power5 chips are used in everything from dual-core OpenPower systems, up through P690 mainframes and Z/os systems. I've also found that getting a cleanly working Sparc64, Linux on Power64, or Linux of IA-64 system to be much easier than on the mixed 32/64-bit opterons. Admittedly, you're right that for most people, 64-bit userland is a hassle, and what you really want is the kernel and certain back-end apps to be 64-bit, but I'm not convinced AMD has the easiest solution there either.
I'd love to see them continue to be in the game, but there are economy of scale issues here which they'll have to address in some manner.
Be careful here. When you talk "64-bit Linux-compatible CPU", you run smack into three problems; IA-64, Power, and SPARC. Admittedly, SPARC really implies Solaris, but the basic point is sound; if AMD were to focus on that market, then it runs flat into three superior and well-established architectures, where it has a toehold with entry-level systems, but only one vendor (Sun) shipping anything of any size (the 4600) involving Opterons. That's probably too small of a market to support the technical innovation necesssary to remain viable, and a good path to oblivion.
Fighting IBM's Power group, and IBM's fabs, doesn't really seem like the best route to success, especially given IBM's committment to Linux on Power. People liked teh Opterons because while they were good chips on their own, they also functioned as a fast Xeon. If you're in a market with Xeonicity doesn't matter, then they're only one option amongst many, and not necessarily the best.
Slashdot Mirror
Ever use a DecPro? Built like tanks, wysiwyg editor with integratable charts and graphics, based on a micro-PDP processor (16-bit flat address space in 1980). Then, some nimrod decided to keep the user locked in this DEC-marketroid approved environment, and that they shouldn't be able to format their own floppies, because DEC would make more money selling pre-formatted ones. They could have just run RSX-11 or RT-11 on it, used the PDP compilers, and instantly had a large installed base with developers everywhere. Instead we got a hacked-together 8-bit processor running a copy of CP/M, and were stuck with that architecture for the next decade and change.
With some vision, they could have been the dominant PC player and become the standard, as they already had a built-in upgrade path, and a decent installed software base. PDP-11 -> VAX -> Alpha Instead, they listened to Ken, marketed wierd machines (still built like tanks) too late (DEC Rainbows), then tried to become a PC company.
What about Wordstar? Glorious program in the early DOS era before WordPervert evolved into a usable product, ran on DOS or CP/M, excellent formatting controls, and didn't need keyboards with either function keys or arrow keys. You could have run it off an ADM3a.
My college, after trying many arguably superior programs (XYWrite, Final Word, Wordstar), mysteriously settled on WordPerfect, probably because a manager thought something driven by func keys that came with a little keyboard template to remind you of what they did was easier. I never understood how a wordprocessor that constantly required you to removed your hands from the home-row keys was supposed to enhance my writing. Then I discovered that nobody can touch-type anyway, so it wasn't making as much difference as I thought.
I'd still probably rather use emacs and TeX, but that wasn't even a dream on early 80s PCs.
Yes, I have, and I still miss them. Your problem was that you (apparently) tried to use Unix on them, rather than VMS, and the common language interface (which allowed you to do system calls and fancy string handling in fortran 77). Once you grokked the Orange Wall (and later Grey Wall), VMS was easy to manage, and rock solid. It used funky networking of course (CMUTEK tcp/ip still gives me shudders), but if you had all VAXes, then DECNET was no big deal. Truly a loss, and superior to many of its successors.
I miss my VAXstation and the 11/785.
turbo pascal 2 was also great, but they never cleanly made the transition to the Windows world. I'm sorry to lose the simplicity of TP2 (which would be great now because you'd just link it to other libraries, rather than rely upon Borland's oddball implementations), and there was always the attempt to be different, such as Turbo Prolog.
Work out at some point what NASA actually spends on science vs the shuttle, and then the alternatives (farm programs, weapons to fight an enemy that no longer exists, office of faith-based initiatives), and you'll see they're not much better off than your field. Coming from an expensive corner of the sciences (chemistry), I used to resent the particle people. 100 authors on a paper, 10s of billions of dollars (and yes, the big ring at Fermilab is way cool, but still...) in construction costs, and a final yield that works out to 1/12th of a fundamental particle/author, if they really saw it. I'm kind of a fan of big telescopes and NASA, as it's the closest most average people come to seeing a frontier and the possibilities therein, and therefore allowing the rest of us to get funded at all. If most people had their way, we'd fund nothing but cancer, fat, and anti-aging research (ok, cynicism off)
Now I'm older and slower, and just sigh and say, "well, at least it's not going to subsidize someone growing rice in a desert".
Propulsion has been solved back in the 60s. The political will to fund the mission and tell the anti-nuke, "let's all go live in unheated mud huts to save the earth" crowd to bugger off is what's lacking. After all, why fund exploration and aspirations when there are farm subsidies and endless petty wars right at hand?
Computers aren't the issue here, mechanical engineering and guts are.
You can get rid of an awful lot of them, though. Despite being stereo only, no home theatre inputs, no remote, my favorite gadget sitting around here is the Harmon-Kardon amp and receiver. Some buttons, and nice big knobs for making adjustments. Tasteful, minimalist readouts. The antithesis of the remote or even front panel on my home theatre receiver.
So when do our troops get the powered armor to go with their nifty new laser cannon?
Of course, maybe it's just time to go for simplicity, and throw rocks from orbit.
Basically, your complaints (and I remember the mysterious "box doing nothing with a load of 25" issue" (probably X related), and the replies (recompile everything from scratch after patching), point towards why it's time for you to try *BSD or more likely, Solaris 10. If your boss will spring for new hardware, of course, you should just get a copy of OSX Server with unlimited client access. 15 years ago in Grad school my boss watched me IPL a VMS box from tape, read the console for a few minute, and then emphatically told me that machines like that were supposed to be extinct by now. One might argue the same for OS's which still require you to manually configure the source, and then start recompiling the kernel and libraries. I'm running my lab off an OSX server box, with all the standard services (it manages both the desktops and the compute nodes on the private network), and really would be hard pressed to give a reason that I thought I had to change to Linux, Windows, or anything else. Nobody has ever suggested I recompile anything on that box to make it work right, and there's very little you can't tweak from the gui or through editing rc.local (good old BSD one file for tuning).
A few years back, when Sun was pushing its Linux desktop, it made its sales reps (at least the ones in the upper midwest), use it on their business laptops. Within the year, every one of those laptops had somehow mutated into PowerBooks running OSX, with far happier reps using them. I somehow transitioned from a dual xeon to an old G4 without noticing, because of the apps and seamless experience which just let me work, which is what I suspect you've seen with your Powerbook as well. Someome took the time to add some polish, and it shows.
No, Linux is a hobby. If you ran a real unix (Solaris, AIX), you'd find that same or better stability than you've ever seen with Windows. You'd find better stability yet with Windows if they'd kept Cutler's VMS derived architecture, and didn't shoe-horn home-user Windows features on top of it for compatibility and convenience.
Nothing personal, but the only Vorbis I worry about is the Grand Exquisitor.
Outside slashdot, there is really only one DAP format: MP3, despite the attempts of Apple, Microsoft, and Real to convince people otherwise.
That's true, because if Clippy was true AI, he'd learn from you and become more helpful. OTOH, he'd also have instructions from MS hard-coded in for features that are detrimental to you, but good for their stock price. The end result would be self-confident schizo app (c.f. HAL 9000), who would probably do Bad Things to your documents at an inopportune time.
At 256 M, you're walking (or dragging) XP, not running it. That being said, I've used it on a 400MHz/228 M laptop, and once it's running, it's not too bad. Not desirable, but not bad.
Not to mention the inevitable driver issues. The bad one we hit was certain Trident cards, all claiming to be the same chip, weren't. Minor revisions resulted in different appearances (or working at all) on nominally identical cards bought as a batch. My suggestion to buy ATI or Tseng-based cards only didn't go over well, as it wasn't the cheapest solution.
Yes, but they made such a prettier chime when you rebooted than the Windows box.
Another vote for OS/2. I ran 2.1, then the v3.0 betas through the 4 series. Wrote my thesis using DeScribe on a 386/40. I see you can get Firefox for it; maybe I'll put it on an old Thinkpad for nostalgia, though I'm not certain about wireless support.
Like BeOS, another step into the future, doomed by poor marketing and fighting a pre-installed based of DOS/Windows.
Yes. Making a 16-bit call was referred to, as I remember, "Thunking", probably from the sound your machine made while it did it.
Dragging to the Applications folder was too hard? Sounds like the old Far Side cartoon with two polar bears, an eskimo running off into the distance, and an igloo they just lifted up: "I lift, you grab. Was that just a little too complicated for you Carl?"
I also presume in this case Unintuitive = "Not Gnome".
YMMV, but I ran unices since the 80s, and have gone through everything from TWM (and VWS), through SGI, CDE, and 7 years of Linux on the desktop, including Gnome from 0.3. OSX/Aqua just works. clean, effecient, and doesn't get in the way. Give it another chance, eventually, you'll be assimilated as well, but it's a gentle, California-type assimilation.
I have a Thinkpad A22m from 2000 which is still in active use, so they don't become doorstops after only a few years, if you're sensible. I wouldn't try 3d gaming with it, but given the 1040x1440 screen and a gig of memory, it's more than fine for Word, Photoshop, or web. Laptops are: compact, efficient, quiet, and don't have multiple cables running everywhere. Everything is self-contained, and you can move them if you want to work somewhere else. Now that they come with 17" widescreens for a reasonable price, for most people, there is no reason for an old-style tower or brick design.
Outside of adding another HD or more memory (and even then people take the machine to the shop), nobody outside of geeking circles every opens their desktop system. Given a similar price point, most home users would opt for the space saving design, and simple forklift upgrade when the time came.
NeXTStep isn't dead, it just evolved into OS-X (which given current Mac sales may make it the most prevalent Unix out there). OTOH, you're right about uSoft setting the industry back. In 1987 I could have bought a 386/16 which matched in performance our VAX 11/750 (if you were willing to drop $800 on the math coprocesser). Sun built a Unix workstation around it. Microsoft's answer? 8-bit DOS 5.x. It wasn't until Grad school, when I convinced my boss to buy the Watcom compilers for the lab that we actually made those machines sing. The DOS extender for the Windows 3.0/DOS hold-outs, and native OS/2 32-bit apps, ported straight off the VAX, for everything else.
Then there's the lack of queueing (VMS, late 70s) multiple jobs in the background, and the database-based WinFS that was supposed to ship with Vista (IBM AS/400), Virtualization (IBM, 1960s), TCP/IP (stolen at the last minute from BSD, and relabeled Microsoft TCP/IP), included compiler to encourage tinkering (Unix 1970s, MSDOS 1981). They just have a real bad case of NIH (not invented here), which they should get over.
A former boss who'd dealt professionally with microsoft in the 80s put it, "the problem is, Bill has a vision..... One vision, and it's mired in the 1970s."
btw, the only problem with NeXT is that they were freaking expensive. Mathematica alone must have been $500 of the base price.
Your story about Jim Cordy is similar to the one that was floating around about 10 years back when uSoft put out a unix compatibility layer for NT, and used Korn shell. During the public announcement someone from teh audience pointed out they'd picked one of the more deficient implementations, to which the microsoftie replied in effect, "we believe this version to be just fine". It was then pointed out the critic was David Korn.
Some days I miss the old days, if only because I came from the VAX/VMS side of the fence, and had languages such as VAX Pascal, and VAX FORTRAN, which could be used for systems programming if you wanted (the orange wall had some great examples in Fortran) and were actually readable. I realize that C succeeded in part because it was cheap to free, and came with all Unixes, but it still looks like line-noise to me, and it's still too easy to shoot yourself in the foot with it, not to mention it's barely higher level than some older mini/mainframe assembly languages. Now, because of lazy CSci programs, its successor, C++, is spreading in chemical programming, so I have to deal with mixed C++/Fortran programs. There should just be a rule; if you're doing science, and not writing an OS or games, just use Fortran95. Clean, modern, easy to optimize, talks to libraries such as LAPACK, and relatively safe.
OTOH, I don't miss compiling programs on a 0.6 VUP (~MIPS) shared with 20 other people.
The "not sure personally"... statement gives me pause.
Look, I like (to a point, which is when RMS starts screaming it's GNU/Linux) Linux, it's great for certain applications, and fun to tinker with. However, having just been subjected to two months of intensive Health Care (hospitalized, multiple IV), and spent several of those weeks in a major research hospital (high-tech to the gills, MD/PhD's poking me daily), I really don't want to see something like that deployed until each critical app has been shown to be robust, to communicate well with everything else, and to Not Go Down. My hospital ran my meds and vitals (which are more important than sleep when you're sick, as they'll wake you out of deep REM at 4:00 a.m. to get a blood pressure reading) over a secured wireless system so the nurses had instant access to all the data they needed. They also copied everything on paper and put it in a big, thick, book, which was copied onto the secure optical network. Big instruments (x-ray, etc), had their own embedded, hardened OS.
The issue here is robustness, and it doesn't matter if its SuSE, Win2K server, AIX, or CP/M-86. There is no room whatsoever for political advocacy of an operating environment in a hospital. I'm glad he likes SuSE; I used to use it until it got flaky in the 9.x series, but the OS isn't the issue, the apps are. Show me the apps are stable, and we'll talk about OS advocacy.
We're talking Renaissance operating systems here, so he'll be wearing tall leather boot, a cuirass and morion, with a knife in one hand, and a crossbow in the other. A daVinci-era machine gun is low-efficiency, and requires a horse cart.
Since we can't actually travel to any suspected blackhole/wormhole-entrance, we need the dudes at CERN to whip one up, then someone can step through and hopefully wind up farther away in our own universe.
You first.
Linux has mindshare, and by supporting Linux on lower-end systems, IBM sells businesses Power-based small systems (only available from them), and then gently moves them up the food chain as their needs increase. This way you have IBM 2-core OpenPower systems, and IBM p595s, and eventually (salesmen look off into the distance picturing the bahamas) a Z-series mainframe, all capable of being partitioned to still run the same apps. It's part of their policy of making sure they have a solution for any size of business, or any size problem a business has.
AIX, possibly deservedly, has a reputation as one of the 'odd' Unices. Embracing Linux allowed IBM to appear more interoperable and less threatening to potential customers who remember the days of IBM and the BUNCH.
Actually, I run almost nothing but at this point. OS 10.4 on G5s (I know, the graphical front end is 32-bit, but my apps in the background are mainly 64), and I still have time on a pair of IA-64 systems. Remember, Power is not PPC. PPC is a cut-down version for low-end systems, and IBM was probably happy to get rid of a low-volume, high-demand, customer. Real Power5 chips are used in everything from dual-core OpenPower systems, up through P690 mainframes and Z/os systems. I've also found that getting a cleanly working Sparc64, Linux on Power64, or Linux of IA-64 system to be much easier than on the mixed 32/64-bit opterons. Admittedly, you're right that for most people, 64-bit userland is a hassle, and what you really want is the kernel and certain back-end apps to be 64-bit, but I'm not convinced AMD has the easiest solution there either.
I'd love to see them continue to be in the game, but there are economy of scale issues here which they'll have to address in some manner.
Be careful here. When you talk "64-bit Linux-compatible CPU", you run smack into three problems; IA-64, Power, and SPARC. Admittedly, SPARC really implies Solaris, but the basic point is sound; if AMD were to focus on that market, then it runs flat into three superior and well-established architectures, where it has a toehold with entry-level systems, but only one vendor (Sun) shipping anything of any size (the 4600) involving Opterons. That's probably too small of a market to support the technical innovation necesssary to remain viable, and a good path to oblivion.
Fighting IBM's Power group, and IBM's fabs, doesn't really seem like the best route to success, especially given IBM's committment to Linux on Power. People liked teh Opterons because while they were good chips on their own, they also functioned as a fast Xeon. If you're in a market with Xeonicity doesn't matter, then they're only one option amongst many, and not necessarily the best.