[stolen from Travelling Shoes because I can't be arsed to dig my copy out]
The best commentary on the entire sorry career of Oswald Mosley was issued by P.G. Wodehouse's immortal flibbertigibbet, Bertie Wooster, when he denouced Sir Roderick Spode -- petty demogogue, leader of the fascist "Black Shorts", and would-be dictator -- in The Code of the Woosters:
"The trouble with you, Spode, is that because you have succeeded in inducing a handful of halfwits to disfigure the London scene by going about in black shorts, you think you're someone. You hear them shouting 'Heil, Spode!' and you imagine it is the Voice of the People. That is where you make your bloomer. What the Voice of the People is saying is: 'Look at that frightful ass Spode swanking about in footer bags! Did you ever in your puff see such a perfect perisher?"
180ÂC is considerably hotter than 180F (about 100K hotter, to be absolutist:-))
Point about kosher chicken taken - I'd never really considered it... might have to nip down the local halal butcher and see what there's is like (there being more Muslims than Jews round this particular neck of the woods)
Have a word with the people who wrote the drivers for your devices - they're coding to some godawful old model there, because ever since WDM1.0 (Win98) the driver stack's supported dynamic configuration. This is a gross oversimplification, but it'll do as a first cut:
when you plug a USB device (essentially a serial device) into the USB bus (which is a PnP bus - the same holds true for USB2, FireWire, PCMCIA and any upcoming PnP busses), the function driver gets a hotplug notification telling it its children have changed. It then notifies the PnP manager in turn that *it's* children have changed and the PnP manager in return queries for the list of devices on the bus
once the PnP manager's determined that there's a new device on the bus, it creates a devnode in the device tree - at the moment, the only driver for this devnode is the USB hub driver, plus any bus filter drivers (which we'll ignore for the sake of this discussion). The PnP manager then interrogates the device to find out what it is, and queries the registry to see if there's a driver been installed previously
if necessary, the PnP manager writes a block of device properties to the registry. It then notifies the usermode PnP manager, which tries to locate the function driver and any filter drivers for the device. Once it's located them (which may require them being copied from the installation disk/CD/whatever), the user-mode PnP manager calls back to the kernel PnP manager, telling it to add device drivers, which in turn start the device
Note that with properly implemented drivers (i.e. ones which work as part of the WDM PnP stack rather than some horrendously arcane junk), *no* reboot is required. As a datapoint, none of the devices I've got floating around on my desk at the moment require a reboot on plugging in - the only one I've come across recently was a Nikon LS-30 negative scanner which appeared to have a SCSI implementation from about 1989 (ASPI in full effect...)
So why doesn't it work for you? Well, either a) the USB hub controller silicon on your PC's horrendously horked (possible if it's *very* old - think about 1995-1996, when USB chipsets were still a bit up in the air) or you're running hardware whose vendors haven't bothered learning how to write drivers... after all, did you ever see a laptop user reboot to swap PCMCIA cards? No comfort, I know, but I can assure you that it works for most of the Rest of Us:-)
Wasn't meant to be condescending: apologies if it came over that way...
AIUI, the only people making big Wintel boxes (bit enough to need a NUMA architecture) are Fujitsu, NEC, HPaq, IBM and Unisys - I don't think Wang even exist any more. And NUMA has it's uses: it gives you *many* of the upsides of SMP (ease of development, with a little thought required about your data access patterns) and a lot less pain than breaking your algorithm down into explicitly parallelisable parts, which clusters or vector machines will always require. The NUMA specifications in Windows require 1:3 near:far access timings, which is a hell of a lot better than any of the cheap interconnect technologies, and things like Giganet and Myrinet are eye-wateringly expensive.
It's notable that no-one's running big RDBMSes (big analytics and the like) on Beowolfs - it's just the Wrong Technology for it. Message-passing ain't always the right answer. Oh, and where did you get the "NUMA is implicit message passing" idea from - it needn't be implemented like that and often isn't - there are plenty of shared-memory implementations out there, largely because SMP code can be moved onto shared-memory ccNUMA with very little effort. Migrating a pre-existing shared-memory-SMP application to a message-passing architecture, on the other hand, is heavy juju and doesn't win you much until remote latency >> local latency - 3:1 ain't that
Oh, and my bad - there's no *true* hotswap. Instead you have to use a mirroring memory controller, then you can swap bad sticks out and the mirror'll rebuild. Microsoft *have* got hooks for this and memory hot-add in the OS, though - and if you get servers from the likes of Stratus and Marathon with memory controllers that handle it then you're golden.
The kernel debugger comment was a bit of a (weak) attempt at a joke... though I might find it rather more annoying if I wrote drivers more frequently than once in a blue moon. Anyway, DbgPrint() should be enough for anyone.
Have you tried a version of Windows post 3.1/NT4? Not only does the plug'n'play work automagically *without* the need to modprobe (assuming the hardware supports it), but you've also got a better than even chance that any hardware you go to the shop and buy will Just Work
3 software raid
Been in NT since at least 3.5 - IIRC, before Linux got it (IIRC NT3.5 was released in 1994, which puts us at about kernel 1.0... anyone care to correct me here)
- Support for 64bit address spaces and CPUs. (Where is the ia64 or x86-64 Windows on this?)
Available in Windows - Itanic XP was available in a limited release and 64-bit 2003 Server can be bought pretty much off-the-shelf
- NUMA (Does some version of Windows support Non-Unified Memory arch, may be something from Wang or some other dinosaur company).
Available in Windows. From your comments, you appear unclear as to what NUMA is - it's not "something from Wang or some other dinosaur company", it's a way of architecting non-symmetrical MP hardware. Go read Greg Pfister's "In Search of Clusters", then go and discover how some Big Computers work - try Sequent (now part of IBM), Cray, SGI and Sun for details
- I am sure there are some esoteric network protocols linux supports natively. But I am not so impressed by that.
Now, does Linux support hot-swap RAM? Hot-swap PCI cards? Useable async I/O, even (and no select(3) doesn't count...)? Does it even have a kernel debugger?:-)
Try buying a real (4-6 month old) chicken from a craft butcher rather than some steroid- & antibiotic-addled 8-week old broiler from your local cheap food emporium that's been injected with hydrolysed pork and beef protein so it'll soak up more water. You'll pay a load more for the privilege but in return you'll get a chicken that (shock) tastes of chicken - honestly, they do have flavour:-). Just rub the breast with oil and salt, stick a halved lemon into the body cavity and roast at 180ÂC for an hour and a half.
For our UK readers who don't have a decent butcher nearby, I recommend the turkey and chicken from Kelly Turkey Farms, who raise probably the best chicken and turkey in the world...
Every study ever done on the topic has shown that 1 *nix administrator is able to administer more systems than 1 Windows admin and even though you might pay more for each administrator, you employ fewer people... I'm sure you get the picture
Oh, bullshit. Care to produce a single cite for this wild-ass claim of yours, or is this merely speculation? Because I can produce several studies pointing the other way - see this article for the latest. Sure, you can huff and puff all you want but I'll take concrete data over claims that "PerryMason (535019)", that well-known fount of all knowledge regarding the cost of information systems, has pulled out of his arse.
so long as standards are followed (and work is documented) any reasonable unix admin (and I mean an administrator of any *nix) can pick up how any linux system works
And as for this... sunshine, have you ever worked with a real Unix, rather than the toys you Lunix obsessives dribble on about? If so then tell me what fmt does, and, for a bonus point, why I wouldn't like to use it on HP-UX? And as for Linux having anything more than a nodding acquaintance with standards - why, YUo R TEh FUNN3H! Hell, the different mainstream Linux distros haven't even approached standardisation *with each other* yet, let alone converged with real Unixes.
I don't know about Hellman & Friedman, which sounds like a brand of mayonnaise, but Conrad Black (that's "Lord Black of Crossharbour" to you, mind) is a sometime Canuck turned British citizien - which he did so he could don ermine as the aforementioned Lord.
AFAIk, he's got no ties to the software industry but he's renowned in the UK for being a raving Tory loon (think fiscally libertarian but socially incredibly moralistic - tax is bad, a woman's place is in the home, why don't these darkies know their place etc): he publishes the Daily and Sunday Telegraph and the Spectator (the latter edited by a Tory MP, which kind-of throws the whole impartiality debate into a fairly strong light). IIRC he's responsible for varous US and Ca newspapers through his network of holding companies, mostly called Hollinger something
Great fuckin' idea. Now where's my tank, my nuclear bomb and my aircraft carrier - I'm feelin' like goin' a liberatin', and didn't I pay for all the above goodies? Hey? HEY?
Possibly the reason it doesn't "have broader coverage" is because it's ridiculous?
--
Jon
While I'd agree entirely with your thesis (that it's the ability to access the data, rather than the application per se that's important), there's an additional interesting question arising... let's say that I'm a good little data archivist and store all my documents in an open format (say PDF) to mitigate the risk of my software vendor going out of business. Now, where do I store those PDFs? Are EMC, Hitachi or IBM SANs sufficiently open that I can get the data off them without using EMC/Hitachi/IBM software/firmware? What about the ArcServe or BackupExec storage formats?
You needed to look in the right place: the original flamewar wasn't on Usenet, but on the legendary UNIXHATERS, which was later archived to Usenet/Deja/Google/whatever under the euphonious title list.unix-haters. Something like (unchecked URL ahoy!) http://groups.google.com/groups?q=jargon+eric+grou p:list.unix-haters should yield sufficient tidbits. Failing that, the UNIXHATERS archives are floating around online.
OK: I'll bite. Cut me some slack tho': it's been a while.
This is, I take it, IBM S/390 (specifically MVS/3090) JCL - without the JES bits, which begin/* rather than//. It's got to be fairly recent, 'cos IGYWCG was a fairly recent (late 90s anyway) addition
Line one has is a job statement: the job is named U890898 and is for username ACCT (aren't the single quotes optional unless you've got a space in your username?). I can't remember what the REGION parameter does: I'd hazard a guess that it's all about ensuring you don't grab too much of the LPAR you're running in... The job's to be scheduled as soon as possible, since you've not specified any TIMING or CLASS parameters and I *think* you're also dumping all the messages, since you've not specified MSGLEVEL or MSGCLASS parameters either - unless the operators have a default message setup for your LPAR
Line 2 is an execute statement: in step 1 of the job, we are to execute the catalogued procedure IGYWCG (this could more properly be written as//STEP1 EXEC PROC=IGYWCG) - compile and run the COBOL program. Now, this is where it all gets a little sketchy, because I'm not sure what the PARM set you've passed is: I assume CPARM is equivalent to PARM.COB but the PARM data don't ring any bells.
The original authors believe this has already happened.
(go and check the Unix-Haters' archives - IIRC then somewhere in there there's a heated thread with ESR attempting to defend what he did to the original MIT JARGON, and the SAIL equivalent, to its original authors - basically ESR had rewritten all the jokes at Unix's expense so they became jokes at MS-DOS's expense, thus causing some to feel that he'd missed the point)
There are if you're a corporate: Orange offers flat-rate pricing for businesses, including Orange GPRS Business LAN, which is billed by the size of the pipe from their NOC to yours...
A poll done for The Sun (a Murdoch-owned and hence very pro-war British tabloid) by MORI indicated that if "UN inspectors do not find proof that Iraq is trying to hide weapons of mass destruction, and the UN security council does not vote in favour of military action", 63% opposed military action and a further 11% were undecided. See MORI's website for the full details. I don't know if the Sun published these results...
How many others are known by name? Nagoshi? Suzuki? Kojima? Spector? Garriott?
Well, from a generation earlier, you forgot Eugene Jarvis (Defender, Joust), Roberta Williams (early Sierra adventures), Dave Theurer (Missile Attack, Tempest), David Braben & Ian Bell (Elite). More recently, Sid Meier (Civilisation), Will Wright (Sim City & The Sims) and Peter Molineaux (all the other god games). And with almost fifty years in the industry between them, Jeff Minter and Mark Cerny. And that's just off the top of my head;-)
But yeah, there's a long and inglorious history of the video game industry ignoring its auteurs. Dates all the way back to Atari, who refused to allow game creators to be named.
Out of interest, how does component reuse come under the heading of "wierd things regarding the API". It looks to me like the developer of RhymBox (your example above) is doing precisely what he or she should: using pre-provided functionality rather than rolling his or her own DHTML renderer.
Not that I'm denying that many developers write (or at least used to write) apps which play fast and loose with the Win32 API - hell, the amount of code which plain wouldn't run on NT is indicative enough of that - but I don't think your example's particularly relevant.
(Disclaimer: I am in no way associated with anyone having anything to do with RhymBox, although I do use both COM and messaging applications:-))
MS-OS/2 6.0 probably would have been a better OS than MS-WXP
Debateable: by now, MS would have been swallowed up by IBM and IBM, as was seen at the time, Didn't Get It when it came to PCs (in IBM's world-view, your PC was merely a smart terminal for connecting to your leased-from-IBM mainframe). You might as well say that the Hurd probably would have been a better OS than Linux, or that TOPS-10 would have been a better OS than VMS or...
You mean the DOS torture box? No one expected it to work
Except those people who listened to IBM's sales force (the white-shirted, navy-blue-tied ones) explaining how "your investment in software for IBM's PC-DOS is safe because it all runs on IBM's new advanced OS/2" (or words to that effect: I can't be arsed to go digging out ancient marketing bumf)
> As long as you weren't waiting for applications which ran on your shiny new requires-4MB-of-RAM-costing-to-boot OS/2?
I could not quite make your phrase, but I assume you are talking about performance and resources requirement. I was speaking about the speed at which OS/2 itself was developed. But even so, OS/2 proved to run better and in less memory than MS-WNT.
I was actually referring to the complete dearth of applications from ISVs for OS/2 - at the time when OS/2 had a chance in the marketplace (i.e. before the launch of Windows 3.0) you could a) single-task DOS applications in MS-DOS/PC-DOS or the DR-DOS equivalents b) multi-task the same DOS applications using QuarterDeck's (very funky) DesqView or c) multi-task the same DOS applications using OS/2 - an exercise which was both far flakier and far more expensive (due to the fact that you realistically needed at least 4MB RAM to even start playing and 6MB to get serious multitasking done. At the time, that was the thick end of $2500 on RAM alone. Add on $700 for an OS/2 licence and it became a very losing proposition). The killer OS/2 app was going to be, as I alluded, OfficeVision which would run on everything from PS/2s running OS/2 right up to MVS in a VM on the mainframe, would integrate with systems like SNADS, DISOSS and PROFS and generally do everything short of make dinner and polish the furniture. IIRC, it never happened (or if it did, it was so late and made so little impact on the market that I never noticed). Think of it as the Visi-On of the late 80s (remember Visi-On? People think MS are the arch-FUDsters? Heh - how little they remember...)
MS attempts at marketing OS/2 were feeble at best
That's not entirely true: cast your mind back to the late 80s, when Windows was still tiled... At the time, Microsoft was pushing OS/2 fairly hard, and Windows was the red-headed stepchild (and Word? Word was a Mac application, and wouldn't make it to the PC world for another three or four years). It wasn't until the 89-90 timeframe, when BillG realised that IBM were about to eat Microsoft's lunch by turning them from a (reasonably successful) DOS vendor into a spanner shop which would grind out OS/2 code that he went off the idea.
Certainly, Microsoft pushed OS/2 harder than they did DOS
If you call the AS/400 expensive or compare it to PCs you never went near one. At any rate, never understood one
Should I show you my RPG/400 scars:-) I explained myself badly here - I'll try again. At the time, IBM were selling AS/400s starting at about $1/4 million. It was what your SME bought, before they grew enough to need an IBM mainframe. And at the time, the AS/400, while a great box in many ways - wierd-ass architecture and all - wasn't exactly a computing powerhouse. The Small Systems division of IBM were very worried that a (relatively) commodity 386 machine would make them look very uncompetitive on a price/performance basis: who'd buy into DB/400 - OS/400 - AS/400 when then could buy Informix or Oracle on SCO Xenix on a real 32-bit machine at a tenth the price? And once they'd strayed from the IBM fold, how would they be brought back? And if that happened, Armonk wouldn't be happy. So the SS division pulled VPs on the Personal Computers and Personal Systems divisions and got OS/2 crippled so as to avoid cannibalising *their* market share.
This was all recounted to me by a friend who used to work at IBM (from mid 70s to early 90s - he was there during the great days of Big Blue middle management politics) so take it with a pinch of salt... but not too big a pinch - he was in mainframes so had no axe to grind.
Are you malicious or just ignorant?
Neither: I'd suggest that I'm better informed (or at least less blinkered) than you...
IBM OS/2 also has interoperability with Unix and the Net, and a better POSIX compatibility history than MS-WNT
Yes. Since OS/2 3.0 Warp. By which time it had already lost. As I said above, back when it had a chance it had SAA written all over it (OS/2 1.1, which was the first version which even vaguely worked, had as its big selling point for the Extended Edition the fact that you could hang it off your Channel really easily - hell, MicroChannel Architecture was a reference to the real Channel, implying that the bus inside your peecee was equivalent to the one that allowed you to hang thousands of 3270s and 5250s off your 370 box...). Again, back in the timeframe we're discussing, TCP/IP and "the Net" were irrelevancies. You're also forgetting that one of the big advantages of SAA/CA (in IBM's eyes) was that it locked the user in: once you'd invested huge amount of time and effort in learning one SAA/CA application, the rest were cheap, but moving over to DEC or PC LANs would be an expensive proposition.
> As long as you were prepared to fork over $3000 for said documentation What is actually quite reasonable for print documentation for an ISV or third-party developer
No it wasn't: if you wanted to start writing code for OS/2, you needed a copy of the OS/2 SDK, and any other tools. Cost: $3000, plus the price of your tools. If you wanted to start writing code for Windows, you needed a copy of the Windows SDK, and any other tools. Cost: 25 cents for a stamp, plus the price of your tools. Microsoft were handing out documentation binders and disks like sweeties at just about any trade fair you went to at the time - if you played your cards right, and asked nicely, they'd slip you a copy of MSC 5 as well, with the wonderful Programmer's Whipping Bench. All you needed to do was buy a copy of Petzold.
> you never used OS/2, did you Yes, I did: versions 1.3, 2.0, 2.1 and 3. It was rock solid, provided you had enough system resources
I was being a bit sarcastic here: you've got all the outward appearances of a Team OS/2 member, and we know what that means:-). Oh, and early OS/2 wasn't rock solid at all: you could take a fully loaded System 80, tricked out with 8MB RAM, load the Presentation Manager from OS/2 1.3 onto it, start up a couple of apps and watch it fall over. Repeatedly. Hell, you could get it to fall over with just the Presentation Manager running. It may have got better later - who cared? Windows was good enough - crashed just about as often, cost about a third as much and ran DOS applications better.
To summarise my observations: when OS/2 had a chance in the marketplace, in the late 80s, IBM blew it by making it a) too expensive, b) too greedy for hardware, c) too much of a blatant attempt at vendor lockin, d) too buggy e) having no software f) being sabotaged by internal competition and g) having no compelling advantages to counter the above. By the time (in about 92 - 93) it overcame these disadvantages it had become an irrelevance. Being "a better Windows than Windows", gaining a TCP/IP stack, having the (admittedly very sweet) Workplace Shell and so on all boiled down to rearranging the deckchairs on the Titanic: the battle had already been lost.
Well, there's no option but to put the video drivers into the kernel since they need direct hardware access and you can't get that from userland. It's also worth noting that the GUI per se is all in userland (it's better known as explorer.exe...). What happened in NT4 is that a large chunk of GDI (the abstraction layer for graphics primitives and surfaces) moved from userland into the kernel. Since chunks of GDI are (or at least can be) provided by the hardware manufacturers - modern video cards have hardware acceleration for GDI primitives such as "draw a line from here to here" - you've suddenly a) added an additional large chunk of code into your kernel and b) added a bunch of other people's code into the kernel - both potentially bad things.
I would love to see an OS take advantage of more than 2 of the 4 "rings" an x86 processor has
IIRC the reason NT doesn't do this is because not all the hardware they originally targetted had more than two rings - the MIPS chips (I think) were particular offenders here. With regard to the GDI being in the kernel: anyone that's running an NT-based OS as a server runs it with the generic VGA driver rather than whatever dreck the chipset manufacturer's shovelled out of the door...
IBM didn't need Microsoft to stab OS/2 in the back: Big Blue did a perfectly good job of that itself.
IBM quality
You never tried running DOS apps in OS/2 1.0, did you. To say it was buggy was an understatement...
OS/2 development speed wasn't so bad
As long as you weren't waiting for applications which ran on your shiny new requires-4MB-of-RAM-costing-to-boot OS/2? Whatever happened to OfficeVision again;-)
The real problem is that MS wanted to push forward the stupid MS Win16 API inherited from MS Windows 1.0 for the IBM-PC/XT
No, the real problem was that the marketplace didn't want OS/2: even when Microsoft were fully committed to it (the period thru' mid-89) and pushing it as the best thing since sliced bread, no-one (outside IBM fiefdoms) bought it. If you want to blame anyone for this state of affairs, blame the clone makers who had no interest in handing control of the PC marketplace back to IBM. IIRC, at the time and with all of DOS's flaws, and with both IBM and Microsoft telling the market that DOS was dead and OS/2 was the way forward, DOS was outselling OS/2 by a ratio of about 200:1.
Oh, and targetting the 286 was a bad idea from the get go, but IBM didn't want to eat into sales of their (expensive) AS/400 line.
while IBM would have made it interoperable
... as part of SAA - so it would have interoperated with... ooh, let's see... that'd be the IBM AS/400 minicomputer and the IBM S/370 mainframe and... do we all get the idea here?
documented
As long as you were prepared to fork over $3000 for said documentation
and stable
I get it: you never used OS/2, did you. Because if you did, you'd know perfectly well that when OS/2 shipped it was so flakey it made Windows Me look like a beacon of system stability. Maybe by OS/2 3.0 it had become moderately usable, but OS/2 1.0, 1.1, 1.2 and 1.3 sucked in so many different ways and had uptimes measured in minutes. Datapoint: a customer at the time had about 500 PS/2 Model 60s running OS/2 1.3 w/ Presentation Manager, one in each store. At the time, this was an investment, in hardware alone of about $3 million, plus about another third of a million in OS licences (yes, only ten years ago $700 was considered a very reasonable pricepoint for a desktop operating system licence - if you were IBM...) They were shockingly unusable - as in the store manager would be entering that day's sales into a dBase app and running the numbers and they'd be able to enter stuff successfully about one time in three - the other two, OS/2 or one of the apps running under it would crap itself, lock the queue and once that happened it was goodnight Vienna.
Slashdot Mirror
The best commentary on the entire sorry career of Oswald Mosley was issued by P.G. Wodehouse's immortal flibbertigibbet, Bertie Wooster, when he denouced Sir Roderick Spode -- petty demogogue, leader of the fascist "Black Shorts", and would-be dictator -- in The Code of the Woosters:
"The trouble with you, Spode, is that because you have succeeded in inducing a handful of halfwits to disfigure the London scene by going about in black shorts, you think you're someone. You hear them shouting 'Heil, Spode!' and you imagine it is the Voice of the People. That is where you make your bloomer. What the Voice of the People is saying is: 'Look at that frightful ass Spode swanking about in footer bags! Did you ever in your puff see such a perfect perisher?"
Point about kosher chicken taken - I'd never really considered it... might have to nip down the local halal butcher and see what there's is like (there being more Muslims than Jews round this particular neck of the woods)
Note that with properly implemented drivers (i.e. ones which work as part of the WDM PnP stack rather than some horrendously arcane junk), *no* reboot is required. As a datapoint, none of the devices I've got floating around on my desk at the moment require a reboot on plugging in - the only one I've come across recently was a Nikon LS-30 negative scanner which appeared to have a SCSI implementation from about 1989 (ASPI in full effect...)
So why doesn't it work for you? Well, either a) the USB hub controller silicon on your PC's horrendously horked (possible if it's *very* old - think about 1995-1996, when USB chipsets were still a bit up in the air) or you're running hardware whose vendors haven't bothered learning how to write drivers... after all, did you ever see a laptop user reboot to swap PCMCIA cards? No comfort, I know, but I can assure you that it works for most of the Rest of Us :-)
AIUI, the only people making big Wintel boxes (bit enough to need a NUMA architecture) are Fujitsu, NEC, HPaq, IBM and Unisys - I don't think Wang even exist any more. And NUMA has it's uses: it gives you *many* of the upsides of SMP (ease of development, with a little thought required about your data access patterns) and a lot less pain than breaking your algorithm down into explicitly parallelisable parts, which clusters or vector machines will always require. The NUMA specifications in Windows require 1:3 near:far access timings, which is a hell of a lot better than any of the cheap interconnect technologies, and things like Giganet and Myrinet are eye-wateringly expensive.
It's notable that no-one's running big RDBMSes (big analytics and the like) on Beowolfs - it's just the Wrong Technology for it. Message-passing ain't always the right answer. Oh, and where did you get the "NUMA is implicit message passing" idea from - it needn't be implemented like that and often isn't - there are plenty of shared-memory implementations out there, largely because SMP code can be moved onto shared-memory ccNUMA with very little effort. Migrating a pre-existing shared-memory-SMP application to a message-passing architecture, on the other hand, is heavy juju and doesn't win you much until remote latency >> local latency - 3:1 ain't that
Oh, and my bad - there's no *true* hotswap. Instead you have to use a mirroring memory controller, then you can swap bad sticks out and the mirror'll rebuild. Microsoft *have* got hooks for this and memory hot-add in the OS, though - and if you get servers from the likes of Stratus and Marathon with memory controllers that handle it then you're golden.
The kernel debugger comment was a bit of a (weak) attempt at a joke... though I might find it rather more annoying if I wrote drivers more frequently than once in a blue moon. Anyway, DbgPrint() should be enough for anyone.
Have you tried a version of Windows post 3.1/NT4? Not only does the plug'n'play work automagically *without* the need to modprobe (assuming the hardware supports it), but you've also got a better than even chance that any hardware you go to the shop and buy will Just Work
3 software raid
Been in NT since at least 3.5 - IIRC, before Linux got it (IIRC NT3.5 was released in 1994, which puts us at about kernel 1.0... anyone care to correct me here)
Available in Windows
- Support for 64bit address spaces and CPUs. (Where is the ia64 or x86-64 Windows on this?)
Available in Windows - Itanic XP was available in a limited release and 64-bit 2003 Server can be bought pretty much off-the-shelf
- NUMA (Does some version of Windows support Non-Unified Memory arch, may be something from Wang or some other dinosaur company).
Available in Windows. From your comments, you appear unclear as to what NUMA is - it's not "something from Wang or some other dinosaur company", it's a way of architecting non-symmetrical MP hardware. Go read Greg Pfister's "In Search of Clusters", then go and discover how some Big Computers work - try Sequent (now part of IBM), Cray, SGI and Sun for details
- I am sure there are some esoteric network protocols linux supports natively. But I am not so impressed by that.
Yeah, Windows has those too...
Now, does Linux support hot-swap RAM? Hot-swap PCI cards? Useable async I/O, even (and no select(3) doesn't count...)? Does it even have a kernel debugger? :-)
For our UK readers who don't have a decent butcher nearby, I recommend the turkey and chicken from Kelly Turkey Farms, who raise probably the best chicken and turkey in the world...
Oh, bullshit. Care to produce a single cite for this wild-ass claim of yours, or is this merely speculation? Because I can produce several studies pointing the other way - see this article for the latest. Sure, you can huff and puff all you want but I'll take concrete data over claims that "PerryMason (535019)", that well-known fount of all knowledge regarding the cost of information systems, has pulled out of his arse.
so long as standards are followed (and work is documented) any reasonable unix admin (and I mean an administrator of any *nix) can pick up how any linux system works
And as for this... sunshine, have you ever worked with a real Unix, rather than the toys you Lunix obsessives dribble on about? If so then tell me what fmt does, and, for a bonus point, why I wouldn't like to use it on HP-UX? And as for Linux having anything more than a nodding acquaintance with standards - why, YUo R TEh FUNN3H! Hell, the different mainstream Linux distros haven't even approached standardisation *with each other* yet, let alone converged with real Unixes.
AFAIk, he's got no ties to the software industry but he's renowned in the UK for being a raving Tory loon (think fiscally libertarian but socially incredibly moralistic - tax is bad, a woman's place is in the home, why don't these darkies know their place etc): he publishes the Daily and Sunday Telegraph and the Spectator (the latter edited by a Tory MP, which kind-of throws the whole impartiality debate into a fairly strong light). IIRC he's responsible for varous US and Ca newspapers through his network of holding companies, mostly called Hollinger something
Possibly the reason it doesn't "have broader coverage" is because it's ridiculous? -- Jon
Just a thought...
-- Jon
This is, I take it, IBM S/390 (specifically MVS/3090) JCL - without the JES bits, which begin /* rather than //. It's got to be fairly recent, 'cos IGYWCG was a fairly recent (late 90s anyway) addition
Line one has is a job statement: the job is named U890898 and is for username ACCT (aren't the single quotes optional unless you've got a space in your username?). I can't remember what the REGION parameter does: I'd hazard a guess that it's all about ensuring you don't grab too much of the LPAR you're running in... The job's to be scheduled as soon as possible, since you've not specified any TIMING or CLASS parameters and I *think* you're also dumping all the messages, since you've not specified MSGLEVEL or MSGCLASS parameters either - unless the operators have a default message setup for your LPAR
Line 2 is an execute statement: in step 1 of the job, we are to execute the catalogued procedure IGYWCG (this could more properly be written as //STEP1 EXEC PROC=IGYWCG) - compile and run the COBOL program. Now, this is where it all gets a little sketchy, because I'm not sure what the PARM set you've passed is: I assume CPARM is equivalent to PARM.COB but the PARM data don't ring any bells.
Do I get partial credit?
(go and check the Unix-Haters' archives - IIRC then somewhere in there there's a heated thread with ESR attempting to defend what he did to the original MIT JARGON, and the SAIL equivalent, to its original authors - basically ESR had rewritten all the jokes at Unix's expense so they became jokes at MS-DOS's expense, thus causing some to feel that he'd missed the point)
There are if you're a corporate: Orange offers flat-rate pricing for businesses, including Orange GPRS Business LAN, which is billed by the size of the pipe from their NOC to yours...
A poll done for The Sun (a Murdoch-owned and hence very pro-war British tabloid) by MORI indicated that if "UN inspectors do not find proof that Iraq is trying to hide weapons of mass destruction, and the UN security council does not vote in favour of military action", 63% opposed military action and a further 11% were undecided. See MORI's website for the full details. I don't know if the Sun published these results...
Bombing Iraqi women and children, for starters </flamebait>
Well, from a generation earlier, you forgot Eugene Jarvis (Defender, Joust), Roberta Williams (early Sierra adventures), Dave Theurer (Missile Attack, Tempest), David Braben & Ian Bell (Elite). More recently, Sid Meier (Civilisation), Will Wright (Sim City & The Sims) and Peter Molineaux (all the other god games). And with almost fifty years in the industry between them, Jeff Minter and Mark Cerny. And that's just off the top of my head ;-)
But yeah, there's a long and inglorious history of the video game industry ignoring its auteurs. Dates all the way back to Atari, who refused to allow game creators to be named.
So I take it you've never been exposed to RPG/400 or, for that matter, JCL. Boy, do you have some fun coming :-)
Not that I'm denying that many developers write (or at least used to write) apps which play fast and loose with the Win32 API - hell, the amount of code which plain wouldn't run on NT is indicative enough of that - but I don't think your example's particularly relevant.
(Disclaimer: I am in no way associated with anyone having anything to do with RhymBox, although I do use both COM and messaging applications :-))
Debateable: by now, MS would have been swallowed up by IBM and IBM, as was seen at the time, Didn't Get It when it came to PCs (in IBM's world-view, your PC was merely a smart terminal for connecting to your leased-from-IBM mainframe). You might as well say that the Hurd probably would have been a better OS than Linux, or that TOPS-10 would have been a better OS than VMS or...
You mean the DOS torture box? No one expected it to work
Except those people who listened to IBM's sales force (the white-shirted, navy-blue-tied ones) explaining how "your investment in software for IBM's PC-DOS is safe because it all runs on IBM's new advanced OS/2" (or words to that effect: I can't be arsed to go digging out ancient marketing bumf)
> As long as you weren't waiting for applications which ran on your shiny new requires-4MB-of-RAM-costing-to-boot OS/2?
I could not quite make your phrase, but I assume you are talking about performance and resources requirement. I was speaking about the speed at which OS/2 itself was developed. But even so, OS/2 proved to run better and in less memory than MS-WNT.
I was actually referring to the complete dearth of applications from ISVs for OS/2 - at the time when OS/2 had a chance in the marketplace (i.e. before the launch of Windows 3.0) you could a) single-task DOS applications in MS-DOS/PC-DOS or the DR-DOS equivalents b) multi-task the same DOS applications using QuarterDeck's (very funky) DesqView or c) multi-task the same DOS applications using OS/2 - an exercise which was both far flakier and far more expensive (due to the fact that you realistically needed at least 4MB RAM to even start playing and 6MB to get serious multitasking done. At the time, that was the thick end of $2500 on RAM alone. Add on $700 for an OS/2 licence and it became a very losing proposition). The killer OS/2 app was going to be, as I alluded, OfficeVision which would run on everything from PS/2s running OS/2 right up to MVS in a VM on the mainframe, would integrate with systems like SNADS, DISOSS and PROFS and generally do everything short of make dinner and polish the furniture. IIRC, it never happened (or if it did, it was so late and made so little impact on the market that I never noticed). Think of it as the Visi-On of the late 80s (remember Visi-On? People think MS are the arch-FUDsters? Heh - how little they remember...)
MS attempts at marketing OS/2 were feeble at best
That's not entirely true: cast your mind back to the late 80s, when Windows was still tiled... At the time, Microsoft was pushing OS/2 fairly hard, and Windows was the red-headed stepchild (and Word? Word was a Mac application, and wouldn't make it to the PC world for another three or four years). It wasn't until the 89-90 timeframe, when BillG realised that IBM were about to eat Microsoft's lunch by turning them from a (reasonably successful) DOS vendor into a spanner shop which would grind out OS/2 code that he went off the idea.
Certainly, Microsoft pushed OS/2 harder than they did DOS
If you call the AS/400 expensive or compare it to PCs you never went near one. At any rate, never understood one
Should I show you my RPG/400 scars :-) I explained myself badly here - I'll try again. At the time, IBM were selling AS/400s starting at about $1/4 million. It was what your SME bought, before they grew enough to need an IBM mainframe. And at the time, the AS/400, while a great box in many ways - wierd-ass architecture and all - wasn't exactly a computing powerhouse. The Small Systems division of IBM were very worried that a (relatively) commodity 386 machine would make them look very uncompetitive on a price/performance basis: who'd buy into DB/400 - OS/400 - AS/400 when then could buy Informix or Oracle on SCO Xenix on a real 32-bit machine at a tenth the price? And once they'd strayed from the IBM fold, how would they be brought back? And if that happened, Armonk wouldn't be happy. So the SS division pulled VPs on the Personal Computers and Personal Systems divisions and got OS/2 crippled so as to avoid cannibalising *their* market share.
This was all recounted to me by a friend who used to work at IBM (from mid 70s to early 90s - he was there during the great days of Big Blue middle management politics) so take it with a pinch of salt... but not too big a pinch - he was in mainframes so had no axe to grind.
Are you malicious or just ignorant?
Neither: I'd suggest that I'm better informed (or at least less blinkered) than you...
IBM OS/2 also has interoperability with Unix and the Net, and a better POSIX compatibility history than MS-WNT
Yes. Since OS/2 3.0 Warp. By which time it had already lost. As I said above, back when it had a chance it had SAA written all over it (OS/2 1.1, which was the first version which even vaguely worked, had as its big selling point for the Extended Edition the fact that you could hang it off your Channel really easily - hell, MicroChannel Architecture was a reference to the real Channel, implying that the bus inside your peecee was equivalent to the one that allowed you to hang thousands of 3270s and 5250s off your 370 box...). Again, back in the timeframe we're discussing, TCP/IP and "the Net" were irrelevancies. You're also forgetting that one of the big advantages of SAA/CA (in IBM's eyes) was that it locked the user in: once you'd invested huge amount of time and effort in learning one SAA/CA application, the rest were cheap, but moving over to DEC or PC LANs would be an expensive proposition.
> As long as you were prepared to fork over $3000 for said documentation
What is actually quite reasonable for print documentation for an ISV or third-party developer
No it wasn't: if you wanted to start writing code for OS/2, you needed a copy of the OS/2 SDK, and any other tools. Cost: $3000, plus the price of your tools. If you wanted to start writing code for Windows, you needed a copy of the Windows SDK, and any other tools. Cost: 25 cents for a stamp, plus the price of your tools. Microsoft were handing out documentation binders and disks like sweeties at just about any trade fair you went to at the time - if you played your cards right, and asked nicely, they'd slip you a copy of MSC 5 as well, with the wonderful Programmer's Whipping Bench. All you needed to do was buy a copy of Petzold.
> you never used OS/2, did you
Yes, I did: versions 1.3, 2.0, 2.1 and 3. It was rock solid, provided you had enough system resources
I was being a bit sarcastic here: you've got all the outward appearances of a Team OS/2 member, and we know what that means :-). Oh, and early OS/2 wasn't rock solid at all: you could take a fully loaded System 80, tricked out with 8MB RAM, load the Presentation Manager from OS/2 1.3 onto it, start up a couple of apps and watch it fall over. Repeatedly. Hell, you could get it to fall over with just the Presentation Manager running. It may have got better later - who cared? Windows was good enough - crashed just about as often, cost about a third as much and ran DOS applications better.
To summarise my observations: when OS/2 had a chance in the marketplace, in the late 80s, IBM blew it by making it a) too expensive, b) too greedy for hardware, c) too much of a blatant attempt at vendor lockin, d) too buggy e) having no software f) being sabotaged by internal competition and g) having no compelling advantages to counter the above. By the time (in about 92 - 93) it overcame these disadvantages it had become an irrelevance. Being "a better Windows than Windows", gaining a TCP/IP stack, having the (admittedly very sweet) Workplace Shell and so on all boiled down to rearranging the deckchairs on the Titanic: the battle had already been lost.
Well, there's no option but to put the video drivers into the kernel since they need direct hardware access and you can't get that from userland. It's also worth noting that the GUI per se is all in userland (it's better known as explorer.exe...). What happened in NT4 is that a large chunk of GDI (the abstraction layer for graphics primitives and surfaces) moved from userland into the kernel. Since chunks of GDI are (or at least can be) provided by the hardware manufacturers - modern video cards have hardware acceleration for GDI primitives such as "draw a line from here to here" - you've suddenly a) added an additional large chunk of code into your kernel and b) added a bunch of other people's code into the kernel - both potentially bad things.
IIRC the reason NT doesn't do this is because not all the hardware they originally targetted had more than two rings - the MIPS chips (I think) were particular offenders here. With regard to the GDI being in the kernel: anyone that's running an NT-based OS as a server runs it with the generic VGA driver rather than whatever dreck the chipset manufacturer's shovelled out of the door...
IBM quality
You never tried running DOS apps in OS/2 1.0, did you. To say it was buggy was an understatement...
OS/2 development speed wasn't so bad
As long as you weren't waiting for applications which ran on your shiny new requires-4MB-of-RAM-costing-to-boot OS/2? Whatever happened to OfficeVision again ;-)
The real problem is that MS wanted to push forward the stupid MS Win16 API inherited from MS Windows 1.0 for the IBM-PC/XT
No, the real problem was that the marketplace didn't want OS/2: even when Microsoft were fully committed to it (the period thru' mid-89) and pushing it as the best thing since sliced bread, no-one (outside IBM fiefdoms) bought it. If you want to blame anyone for this state of affairs, blame the clone makers who had no interest in handing control of the PC marketplace back to IBM. IIRC, at the time and with all of DOS's flaws, and with both IBM and Microsoft telling the market that DOS was dead and OS/2 was the way forward, DOS was outselling OS/2 by a ratio of about 200:1.
Oh, and targetting the 286 was a bad idea from the get go, but IBM didn't want to eat into sales of their (expensive) AS/400 line.
while IBM would have made it interoperable
documented
As long as you were prepared to fork over $3000 for said documentation
and stable
I get it: you never used OS/2, did you. Because if you did, you'd know perfectly well that when OS/2 shipped it was so flakey it made Windows Me look like a beacon of system stability. Maybe by OS/2 3.0 it had become moderately usable, but OS/2 1.0, 1.1, 1.2 and 1.3 sucked in so many different ways and had uptimes measured in minutes. Datapoint: a customer at the time had about 500 PS/2 Model 60s running OS/2 1.3 w/ Presentation Manager, one in each store. At the time, this was an investment, in hardware alone of about $3 million, plus about another third of a million in OS licences (yes, only ten years ago $700 was considered a very reasonable pricepoint for a desktop operating system licence - if you were IBM...) They were shockingly unusable - as in the store manager would be entering that day's sales into a dBase app and running the numbers and they'd be able to enter stuff successfully about one time in three - the other two, OS/2 or one of the apps running under it would crap itself, lock the queue and once that happened it was goodnight Vienna.
Also watch out for: reading complicated literature, showing concern for their fellow man and liking to share...