Still, so far, OSX is holding up well on my systems, showing no signs of decay.
For my two cents, Mac OS X may have its own form of decay, but it doesn't seem to affect the operation of the system the way Windows decay does. On my Mac I have quite a bit of cruft-- an installation of PostgreSQL that I no longer use and haven't bothered to disable or remove is a good example. But I never notice it unless I'm looking for disk space.
I took a month-long vacation from work in May. In April, all was well. When I came back at the end of May, suddenly it was taking about fifteen seconds to open Acrobat. Previously, it would pop right open. That's a mystery to me. I've never had anything like that happen on any of my Macs.
I'm disappointed that your post got moderated "insightful." It's not that you're wrong-- as far as I know, you've got your facts right. It's just that you're complaining for the sake of complaining. The fact that Microsoft is the object of your bile makes you "insightful."
Microsoft Windows, out of the box, is the reference standard for a clean operating system. The fact that this isn't the same as jamienk's idea of a clean operating system doesn't change anything.
Consider the opposite scenario. Whenever I install Linux, I blow through the defaults, mostly, just to get the OS up and running. And, every damn time, I have to go back downstairs to the lab and install and chkconfig on the telnet server. How insane is it that Red Hat decided to ship the OS with telnet access off by default? What good is a server if you can't telnet into it? Are they expecting me to sit down in front of the damn thing?
But that doesn't mean Red Hat's default installation is bogus. It just means that it's not completely compatible with my preferences. Same thing here. The default Windows installation isn't compatible with your preferences. So what?
Like I said, I'm just disappointed that you got moderated "insightful" for this comment. If I had mod points today, I'd take you down a point. As I don't, I'll just be satisfied with rebutting your position.
I have to say, you and I agree in practically every way. That's encouraging.
I have no problem with people who release their software under the GPL. They are making their code available under terms that make it absolutely useless to for-profit groups like companies and entrepreneurs, and that's fine. Hell, I even sort of admire the cleverness of it. One can't use GPL'd software without accepting the terms of the license, and those terms dictate that you have to release source code at no charge. That's insidious, and ingenious.
But it's people who attach the word "free" to that idea that really get my goat. GPL'd software is no more free than Microsoft's software is. In each case, the user of the software is bound by the terms of the license agreement. The only difference is in the specific cans-and-can'ts of the licenses.
"Free as in speech" is another one that pisses me off. What the hell does freedom of speech have to do with the GPL? Is it written someplace that, "Congress shall make no law... abridging the freedom to write GPL-licensed software?" Calling GPL-licensed software "free as in speech" is just another rhetorical technique to try to associate the "free" software advocates' message with something the audience considers to be absolutely good-- the Bill of Rights-- and to discourage them from critical examination of what the "free" software advocates actually stand for.
If they-- RMS and his cadre-- would just stand up and use clear and precise language, without all the rhetoric and obfuscation, then I don't think I'd have any problem with them at all. As it is, they're using dishonest techniques to advance an agenda, and that gives me the creeps.
If you read the article you'd know that they're talking about induced rain. That's all. You know there's going to be a soccer match in three days (or whatever), so you put silver iodide in the clouds to make it rain. Get much of the moisture out of the local atmosphere, reduce the likelihood of rain next week. It's a very localized and very well understood process. The hitch, of course, is getting the right amount of stuff into the atmosphere at the right time and in the right way. It's all in the delivery.
Besides, the article is worth it for the headline alone: "Rain called on account of game." LOL.
The problem is not technological, it's a problem of politics and distribution.
You left out "economic."
The problem of starvation is partly, slightly technological. Some foods, particular vegetables, don't travel well, and as such are difficult to transport to the more remote parts of the world. Shipping avocadoes to Ethiopia would be a dicey project at best.
But other foods, of course, can be transported easily. Grains, for example, are practically indestructable, as long as you keep them dry, free of vermin, and not stolen. So it's only slightly and insignificantly a technological problem.
This post of yours is even worse than that one you made about the giant squid.
"...in this brave new world we still know so little about what lurks beneath the indigo waves of the oceans that cover 80% of our planet."
Ah, yes. Purple prose, indeed.;-) Maybe tps12 is just a very specific and thorough crapflooder.
No, I don't think you're being childish. I just think the thing you're choosing to complain about (while "complain" may not be the best word, it's the closet one I can think of to describe what you're doing) is rather silly. Apple obviously didn't call it "Mac OS 10," as in "Mac OS version 10," because it's not Mac OS. It's something new. But on the other hand, the tradition of referring to major releases of the Mac operating system goes back a long way. Remember what a big deal "System 6" was? Or the even bigger deal, "System 7?" They changed the branding a little bit with "Mac OS 8," and continued that pattern with "Mac OS 9." It's clear that Apple wanted to continue the tradition while still separating the old Mac OS from the new operating system. So, "Mac OS X, version 10.0," and "Mac OS X 10.2," and so on.
Whatever your opinion, you have to admit it's better than "Mac OS 2000.";-)
It's not a "naming scheme." It's just a name. "Mac OS X." That's the name. "What operating system are you using?" "I'm using Mac OS X." It's a single product, with a single name.
Version numbers are absolutely arbitrary. The only thing about them that's consistent across all uses is that they increase over time; one may reasonably expect that version n came before version n + m.
Microsoft has stopped using version numbers publicly altogether. It's Windows 2000, with or without various patches. Or it's Windows XP. Does anyone care-- or even know-- that Windows 2000 was referred to internally as Windows NT 5.0, and that Windows XP was Windows NT 5.1?
And you've gotta be kidding with your OS X v. 11/X11 thing. The only people who could possibly be confused by those names are those who have no idea what's being talked about. People who are uninformed will be wrong whether the names are similar or not.
What's so complicated? The name of the product includes a number. The number in the name of the product was initially the same as the version number of the product, but isn't any more. Now the version number of the product is different from the name of the product.
Is that really so complicated? I don't understand why this is a source of cognitive dissonance for so many people. It's "Mac OS X version 10.2." It's not that hard.
If that's the case, then why isn't it version 1.2 instead of 10.2?
Because there was never a version 1.0 of Mac OS X. The first version of Mac OS X was version 10.0. That's easy to understand: the previous version of Mac OS (actually an entirely different product) was 9.0, so the next version (a new product) was called 10.0.
The branding ("Mac OS X") is separate from the version number ("10.2").
While I would feel like a cheap-skate, I would feel vindicated at this outrageous racketeering - $129 for an OS update. I thought only Microsoft (Win98 SE) pulled crap like that.
Read this post. Microsoft and every other OS vendor in the industry charge for feature releases. And all of them charge more for their feature releases than Apple is charging for Jaguar.
The fact that you bought the OS once doesn't mean you're entitled to a free copy of every release of the OS forever. That's a nice idea on its face and all, but it's not in line with industry practices.
Since you're posting as an AC, I have no idea who you are or what your background is. So I'll assume that you're just ignorant, and not stupid.
Free or super-cheap software upgrades are kind of a myth. For example, Microsoft offers upgrades to Windows XP for owners of '98, ME, NT4, and 2000 only, and that price is $199. If you're still running '95, you can only upgrade to XP by buying the full $299 retail package, or by buying a new computer.
At the high end, you typically only get upgrades on operating systems if you buy a support contract. I don't know about Sun or HP or IBM, specifically, but with SGI you have to pay $500 for each point release of the OS, unless you stay under a support contract. (Until recently, it was $2,000 per release.) So to go from 6.5.15f to 6.5.16f, it's $500, and from 6.5.16f to 6.5.17f, it's another $500. And these are minor feature releases, sent out every quarter. They're tiny in comparison to Apple's mostly-annual major feature releases. (SGI has two OS branches: feature [f] and maintenance [m]. You get bug fixes for free within the same major release, but you have to pay for new features. The maintenance releases have replaced the old patch system, where each bug fix was packaged separately and could be downloaded individually.)
So the idea that you should get OS X 10.2 for free or almost for free is out of line with the way the industry works. Bug fixes are free: 10.0.[1-4] and 10.1.[1-5] were free downloads to all users, whether they were under AppleCare or not. Hell, Apple didn't even check to see if you had a pirated copy of OS X; the OS has no serial number mechanism in it at all, so everybody gets bug fixes for free, even if they didn't buy the OS.
And as new feature releases goes, $129 is the lowest price in the industry, as far as I know.
So no, you're wrong. Pricing 10.2 as a for-sale upgrade only (except for specific price-protection situations) won't "sour anyone who bought 10.1 server." Unless they're pretty unreasonable and unrealistic people with no knowledge of how this sort of thing usually works, they won't be "soured" at all.
It's important to distinguish between "OS X" and "10.n." The name of the operating system is "Mac OS X," pronounced "Mac oh-ess ten." That's the name, just "Solaris" is the name of the operating system.
The version number is (currently) 10.1.5. So it's Mac OS X 10.1.5. In a month, it'll be Mac OS X 10.2. A year from now, it may (but probably won't, by then) be Mac OS X 11.0. There will never be a Mac OS XI, unless Apple decides to change the name of the OS.
Sorry to be so pedantic about this, but I'm just tired of seeing references to "OS X.1.5" and "OS X.2" and "OS XI."
It's not so much stagnation as it is diminishing-returns economics. SGI had plans, back around '98 or '99, to develop a new graphics product code-named Bali that would be superior in most every way to InfiniteReality. The cost of developing that product exceeded the projected customer demand for it. In other words, almost all of the customers could do everything they needed to do with IR, and wouldn't buy Bali. Those customers who would commit to buying Bali couldn't do so in the volumes necessary for SGI to recoup their investment in it.
Unlike most aspects of computing, the demand for high-end graphics capability doesn't appear to be growing faster than the technology. Instead, the demand for highly capable low-end graphics is growing; customers don't want to do anything that they couldn't do with InfiniteReality, but they do want to do it more cheaply. Bali wouldn't have gotten them there. It would have allowed them to do a lot more than IR, but for a lot more money. (Project costs I saw at my company were something on the order of $200,000 per graphics pipe, which is about 25% higher than IR.)
The fundamental issue here is that nobody can think of anything to do with that much power. We can already build real-time photo-realistic simulations with resolutions higher than the visual acuity of the user. What more is there to do, except do that same thing, only cheaper?
Stbility is king, and that is why Linux has entered the enterprise first, not the desktop.
Um... actually, Linux is used on servers more than on desktops because the Linux desktop experience is like having your fingernails pulled out. It's unbelievably awful. I don't like Windows at all, but I'd rather use Windows than Linux on my desktop. At least I can accomplish things with Windows.
It sounds like you don't hate USB so much as you hate USB on Windows.
I use OS X. I a digital camera (a Nikon), and I upload the pictures to my Mac with a USB cable and iPhoto. No drivers required.
I have a USB scanner and a USB printer, and with both of them the story is the same: just plug them in.
There's nothing fundamentally wrong with USB. It sounds, from what you said, that you've ust had really bad experiences because you chose the wrong OS.
Re:not all roses with silicon graphics, er, "sgi"
on
Virtual Sword Fighting
·
· Score: 2
Word on the street says IP isn't even close to the raw performance that Bali would have had....
That's because InfinitePerformance is Odyssey, not Bali. Odyssey is the same one-chip graphics technology used in the Octane2 and the Fuel. (As I'm sure you know; that's just for the less-informed readers out there.) Incidentally, Odyssey is cool. Pull out the board and look at it. What do you see? One chip, about three inches square, with a giant heat sink on it. That's the graphics system. Cool.
I was never in the loop, per se, on Bali, but I did work, at the time, for what was supposed to be one of SGI's big Bali customers. My former company built flight sims, and they were bidding on a project for the US DoD to build the next-generation F-16 trainer. The system was prototyped using a big multipipe IR system (seven racks, just for the IG!), but it was to go into production on Bali.
Long story short, Rayt^W my former employer lost the contract to Lockheed, so all of those pre-orders for Onyx3s with Bali disappeared. I'm sure if that customer hadn't bailed, SGI wouldn't have been so quick to can the Bali project.
I agree with you that desktop graphics have come along way, but I think SGI still has the lead in terms of ability to render a quality visual scene in real time. Pure geometry performance is important, but since the poly count is fairly low in real-time sims anyway, texture and fill performance-- and high-speed texture cache size-- are even more important, and SGI wins those contests hands down.
Well, as I said before, I'm not a graphics programmer-- I'm a different kind of programmer-- so I may get some of these details wrong.
When you say "32 bits per pixel," you're talking about output pixel depth and format. A pixel in RGBA8 format stores one byte for each of red, green, blue, and alpha, and no other data. Those 32 bits are used by the DACs on the hardware to generate a component RGB video signal to drive your monitor. (Or, as I said before, a digital signal, but I'm not familiar with digital signal formats, so I get a little fuzzy at that point.)
IR doesn't support RGB8 or RGBA8; it uses either RGB10 (the default), in which 10 bits are used for each of red, green, and blue (not sure of the packing used), RGBA10 (adds alpha), or RGB12 (12 bpp).
On top of the color data, you can have a second buffer (used to eliminate image flicker in real-time animations), stereoscopic buffers (rendering two different images into the same buffer and display them through special stereo viewing hardware), auxiliary buffers (used for off-screen rendering in hardware; glCopyPixels() can copy aux buffer pixels into the visible frame buffer), multisample antialiasing, Z-buffering, and so on.
As I understand it from my vis sim buddies, it's really not that hard to fill up a 256 bit pixel in a real time image generator. They use 1 Kbit and 2 Kbit pixels pretty often.
Here's an example of a visual available on my Onyx2 at the office:
Why is it sad that SGI designed and built graphics hardware in 1996 that is still technically superior to any graphics system on the market today? I don't think that's sad. I think that's astounding.
I think your definition of "mainstream" is flawed. The software is commercially available. You can buy it right now. I have a system set up in the office down the hall from mine, as a matter of fact. There are hundreds of Inferno systems around the world being used for motion picture and television production. Hell, the next season of Enterprise is being produced in HD, and (starting sometime in the next couple of weeks or so) effects will be created with Infernos at a post production house on Sunset Boulevard.
These products are absolutely, 100% mainstream. The fact that they're expensive doesn't make them not mainstream. The fact that there's no open-source-gimme-gimme version certainly doesn't make them not mainstream.
Combine that with SGI's sudden idea to start selling x86 systems running linux, and I was sure that they were only a year or two away from closing shop.
Yeah, it's interesting that there's a distinct correlation between SGI's dropping all of their x86-based workstations and servers and their being able to actually squeak out a tiny profit one recent quarter.
I guess the lesson there is, "don't build stuff that people won't buy."
You got most of your info about IR4 right, but I just wanted to clarify some things in greater detail. IR is confusing at first, and very different from the typical single-board graphics systems that most people are familiar with. All the details can be found here, but here's an executive summary.
InfiniteReality (be it the original IR on Onyx, or IR2, IR2E, IR3, or now IR4) is comprised of a set of boards. In order to function, the set has to include one geometry engine (or GE), one raster manager (or RM), and one display generator (or DG). The GE board is where the graphics coprocessors live, and it's responsible for most of the 3D math. The DG converts the frame buffer into an analog RGB signal, or a CCIR-601 SD video signal, or, recently, a digital signal.
The RMs are the interesting part. The RM board holds both the frame buffer (80 MB on IR3, 2.5 GB on IR4) and the texture RAM (256 MB on IR3, 1 GB on IR4). A graphics pipe can include one, two, or four raster managers. When you add RMs, you increase frame buffer size (or the size of the raster you can render), but texture cache.
So a four RM graphics pipe will have 10 GB of frame buffer and 1 GB of texture cache, but that 1 GB of texture will be on each of the four RMs. So each texture you download will be stored, in parallel, on each of the four RMs. This keeps texture operations nice and peppy even when you're rendering into a 3840 x 2160 buffer. (That's four times more resolution than HDTV, if you're interested.)
Note, also, that these memories aren't combined. The TRAM and the frame buffer RAM are isolated in hardware. You can't store textures in the frame buffer, and you can't render in texture RAM. So saying that IR4 has a combined 11 GB of graphics RAM is not quite true, and slightly misleading. But only slightly.;-)
The whole thing adds up to an incredibly flexible system. You can configure the graphics pipe as a relatively small raster of 2,048-bit-deep pixels, or an 8-million-pixel raster of 256-bit-deep pixels, or almost anything in between. You can render a truly giant image-- about 3K by 2K pixels, progressive scan, or even more than that if you're willing to live with interlacing-- with full antialiasing, multi-buffered. It's pretty.
(If all you want is pure geometry performance, for viewing giant CAD models and stuff in real time in a VR environment, SGI also has their InfinitePerformance line of graphics hardware for Onyx. But that's another topic.)
Okay, that's enough "Rah-rah, IR" for one night, with just one more little piece of trivia. InfiniteReality graphics has remained fundamentally unchanged since 1996 or so. The only exception is the change from an Everest bus host to an XIO host system. Every few years, SGI has increased the speed of the GEs, or the texture capacity on the RMs, or the performance of the DACs in the DG, but the system itself hasn't really changed at all in six or seven years. That's pretty amazing.
Spoken like a true guy-who's-never-read-the-Aqua-guidelines.
In order to be in line with the Aqua UI guidelines, you have to implement them all, completely. You can't just get kinda-sort close, throw a pinstripe background behind your toolbar and some gummy window decorations, and call it a day.
You should read the Aqua HI guidelines sometime. (Also available in PDF. They just might open your eyes.
BitGeek, I love what you wrote here. I've just finished posting a comment on this subject that I'm pretty pleased with. I suspect that you might agree with the ideas behind it. Would you mind giving it a look?
Slashdot Mirror
How sane is it to have a service that transmits plain text passwords turned on by default?
"It's not whether you're paranoid, Lenny. It's whether you're paranoid enough."
Still, so far, OSX is holding up well on my systems, showing no signs of decay.
For my two cents, Mac OS X may have its own form of decay, but it doesn't seem to affect the operation of the system the way Windows decay does. On my Mac I have quite a bit of cruft-- an installation of PostgreSQL that I no longer use and haven't bothered to disable or remove is a good example. But I never notice it unless I'm looking for disk space.
I took a month-long vacation from work in May. In April, all was well. When I came back at the end of May, suddenly it was taking about fifteen seconds to open Acrobat. Previously, it would pop right open. That's a mystery to me. I've never had anything like that happen on any of my Macs.
I'm disappointed that your post got moderated "insightful." It's not that you're wrong-- as far as I know, you've got your facts right. It's just that you're complaining for the sake of complaining. The fact that Microsoft is the object of your bile makes you "insightful."
Microsoft Windows, out of the box, is the reference standard for a clean operating system. The fact that this isn't the same as jamienk's idea of a clean operating system doesn't change anything.
Consider the opposite scenario. Whenever I install Linux, I blow through the defaults, mostly, just to get the OS up and running. And, every damn time, I have to go back downstairs to the lab and install and chkconfig on the telnet server. How insane is it that Red Hat decided to ship the OS with telnet access off by default? What good is a server if you can't telnet into it? Are they expecting me to sit down in front of the damn thing?
But that doesn't mean Red Hat's default installation is bogus. It just means that it's not completely compatible with my preferences. Same thing here. The default Windows installation isn't compatible with your preferences. So what?
Like I said, I'm just disappointed that you got moderated "insightful" for this comment. If I had mod points today, I'd take you down a point. As I don't, I'll just be satisfied with rebutting your position.
I have to say, you and I agree in practically every way. That's encouraging.
I have no problem with people who release their software under the GPL. They are making their code available under terms that make it absolutely useless to for-profit groups like companies and entrepreneurs, and that's fine. Hell, I even sort of admire the cleverness of it. One can't use GPL'd software without accepting the terms of the license, and those terms dictate that you have to release source code at no charge. That's insidious, and ingenious.
But it's people who attach the word "free" to that idea that really get my goat. GPL'd software is no more free than Microsoft's software is. In each case, the user of the software is bound by the terms of the license agreement. The only difference is in the specific cans-and-can'ts of the licenses.
"Free as in speech" is another one that pisses me off. What the hell does freedom of speech have to do with the GPL? Is it written someplace that, "Congress shall make no law... abridging the freedom to write GPL-licensed software?" Calling GPL-licensed software "free as in speech" is just another rhetorical technique to try to associate the "free" software advocates' message with something the audience considers to be absolutely good-- the Bill of Rights-- and to discourage them from critical examination of what the "free" software advocates actually stand for.
If they-- RMS and his cadre-- would just stand up and use clear and precise language, without all the rhetoric and obfuscation, then I don't think I'd have any problem with them at all. As it is, they're using dishonest techniques to advance an agenda, and that gives me the creeps.
Gee whiz. Fly off the handle much?
If you read the article you'd know that they're talking about induced rain. That's all. You know there's going to be a soccer match in three days (or whatever), so you put silver iodide in the clouds to make it rain. Get much of the moisture out of the local atmosphere, reduce the likelihood of rain next week. It's a very localized and very well understood process. The hitch, of course, is getting the right amount of stuff into the atmosphere at the right time and in the right way. It's all in the delivery.
Besides, the article is worth it for the headline alone: "Rain called on account of game." LOL.
The problem is not technological, it's a problem of politics and distribution.
;-) Maybe tps12 is just a very specific and thorough crapflooder.
You left out "economic."
The problem of starvation is partly, slightly technological. Some foods, particular vegetables, don't travel well, and as such are difficult to transport to the more remote parts of the world. Shipping avocadoes to Ethiopia would be a dicey project at best.
But other foods, of course, can be transported easily. Grains, for example, are practically indestructable, as long as you keep them dry, free of vermin, and not stolen. So it's only slightly and insignificantly a technological problem.
This post of yours is even worse than that one you made about the giant squid.
"...in this brave new world we still know so little about what lurks beneath the indigo waves of the oceans that cover 80% of our planet."
Ah, yes. Purple prose, indeed.
No, I don't think you're being childish. I just think the thing you're choosing to complain about (while "complain" may not be the best word, it's the closet one I can think of to describe what you're doing) is rather silly. Apple obviously didn't call it "Mac OS 10," as in "Mac OS version 10," because it's not Mac OS. It's something new. But on the other hand, the tradition of referring to major releases of the Mac operating system goes back a long way. Remember what a big deal "System 6" was? Or the even bigger deal, "System 7?" They changed the branding a little bit with "Mac OS 8," and continued that pattern with "Mac OS 9." It's clear that Apple wanted to continue the tradition while still separating the old Mac OS from the new operating system. So, "Mac OS X, version 10.0," and "Mac OS X 10.2," and so on.
;-)
Whatever your opinion, you have to admit it's better than "Mac OS 2000."
It's not a "naming scheme." It's just a name. "Mac OS X." That's the name. "What operating system are you using?" "I'm using Mac OS X." It's a single product, with a single name.
Version numbers are absolutely arbitrary. The only thing about them that's consistent across all uses is that they increase over time; one may reasonably expect that version n came before version n + m.
Microsoft has stopped using version numbers publicly altogether. It's Windows 2000, with or without various patches. Or it's Windows XP. Does anyone care-- or even know-- that Windows 2000 was referred to internally as Windows NT 5.0, and that Windows XP was Windows NT 5.1?
And you've gotta be kidding with your OS X v. 11/X11 thing. The only people who could possibly be confused by those names are those who have no idea what's being talked about. People who are uninformed will be wrong whether the names are similar or not.
What's so complicated? The name of the product includes a number. The number in the name of the product was initially the same as the version number of the product, but isn't any more. Now the version number of the product is different from the name of the product.
Is that really so complicated? I don't understand why this is a source of cognitive dissonance for so many people. It's "Mac OS X version 10.2." It's not that hard.
If that's the case, then why isn't it version 1.2 instead of 10.2?
Because there was never a version 1.0 of Mac OS X. The first version of Mac OS X was version 10.0. That's easy to understand: the previous version of Mac OS (actually an entirely different product) was 9.0, so the next version (a new product) was called 10.0.
The branding ("Mac OS X") is separate from the version number ("10.2").
While I would feel like a cheap-skate, I would feel vindicated at this outrageous racketeering - $129 for an OS update. I thought only Microsoft (Win98 SE) pulled crap like that.
Read this post. Microsoft and every other OS vendor in the industry charge for feature releases. And all of them charge more for their feature releases than Apple is charging for Jaguar.
The fact that you bought the OS once doesn't mean you're entitled to a free copy of every release of the OS forever. That's a nice idea on its face and all, but it's not in line with industry practices.
Since you're posting as an AC, I have no idea who you are or what your background is. So I'll assume that you're just ignorant, and not stupid.
Free or super-cheap software upgrades are kind of a myth. For example, Microsoft offers upgrades to Windows XP for owners of '98, ME, NT4, and 2000 only, and that price is $199. If you're still running '95, you can only upgrade to XP by buying the full $299 retail package, or by buying a new computer.
At the high end, you typically only get upgrades on operating systems if you buy a support contract. I don't know about Sun or HP or IBM, specifically, but with SGI you have to pay $500 for each point release of the OS, unless you stay under a support contract. (Until recently, it was $2,000 per release.) So to go from 6.5.15f to 6.5.16f, it's $500, and from 6.5.16f to 6.5.17f, it's another $500. And these are minor feature releases, sent out every quarter. They're tiny in comparison to Apple's mostly-annual major feature releases. (SGI has two OS branches: feature [f] and maintenance [m]. You get bug fixes for free within the same major release, but you have to pay for new features. The maintenance releases have replaced the old patch system, where each bug fix was packaged separately and could be downloaded individually.)
So the idea that you should get OS X 10.2 for free or almost for free is out of line with the way the industry works. Bug fixes are free: 10.0.[1-4] and 10.1.[1-5] were free downloads to all users, whether they were under AppleCare or not. Hell, Apple didn't even check to see if you had a pirated copy of OS X; the OS has no serial number mechanism in it at all, so everybody gets bug fixes for free, even if they didn't buy the OS.
And as new feature releases goes, $129 is the lowest price in the industry, as far as I know.
So no, you're wrong. Pricing 10.2 as a for-sale upgrade only (except for specific price-protection situations) won't "sour anyone who bought 10.1 server." Unless they're pretty unreasonable and unrealistic people with no knowledge of how this sort of thing usually works, they won't be "soured" at all.
It's important to distinguish between "OS X" and "10.n." The name of the operating system is "Mac OS X," pronounced "Mac oh-ess ten." That's the name, just "Solaris" is the name of the operating system.
The version number is (currently) 10.1.5. So it's Mac OS X 10.1.5. In a month, it'll be Mac OS X 10.2. A year from now, it may (but probably won't, by then) be Mac OS X 11.0. There will never be a Mac OS XI, unless Apple decides to change the name of the OS.
Sorry to be so pedantic about this, but I'm just tired of seeing references to "OS X.1.5" and "OS X.2" and "OS XI."
It's not so much stagnation as it is diminishing-returns economics. SGI had plans, back around '98 or '99, to develop a new graphics product code-named Bali that would be superior in most every way to InfiniteReality. The cost of developing that product exceeded the projected customer demand for it. In other words, almost all of the customers could do everything they needed to do with IR, and wouldn't buy Bali. Those customers who would commit to buying Bali couldn't do so in the volumes necessary for SGI to recoup their investment in it.
Unlike most aspects of computing, the demand for high-end graphics capability doesn't appear to be growing faster than the technology. Instead, the demand for highly capable low-end graphics is growing; customers don't want to do anything that they couldn't do with InfiniteReality, but they do want to do it more cheaply. Bali wouldn't have gotten them there. It would have allowed them to do a lot more than IR, but for a lot more money. (Project costs I saw at my company were something on the order of $200,000 per graphics pipe, which is about 25% higher than IR.)
The fundamental issue here is that nobody can think of anything to do with that much power. We can already build real-time photo-realistic simulations with resolutions higher than the visual acuity of the user. What more is there to do, except do that same thing, only cheaper?
Stbility is king, and that is why Linux has entered the enterprise first, not the desktop.
Um... actually, Linux is used on servers more than on desktops because the Linux desktop experience is like having your fingernails pulled out. It's unbelievably awful. I don't like Windows at all, but I'd rather use Windows than Linux on my desktop. At least I can accomplish things with Windows.
It sounds like you don't hate USB so much as you hate USB on Windows.
I use OS X. I a digital camera (a Nikon), and I upload the pictures to my Mac with a USB cable and iPhoto. No drivers required.
I have a USB scanner and a USB printer, and with both of them the story is the same: just plug them in.
There's nothing fundamentally wrong with USB. It sounds, from what you said, that you've ust had really bad experiences because you chose the wrong OS.
Word on the street says IP isn't even close to the raw performance that Bali would have had....
That's because InfinitePerformance is Odyssey, not Bali. Odyssey is the same one-chip graphics technology used in the Octane2 and the Fuel. (As I'm sure you know; that's just for the less-informed readers out there.) Incidentally, Odyssey is cool. Pull out the board and look at it. What do you see? One chip, about three inches square, with a giant heat sink on it. That's the graphics system. Cool.
I was never in the loop, per se, on Bali, but I did work, at the time, for what was supposed to be one of SGI's big Bali customers. My former company built flight sims, and they were bidding on a project for the US DoD to build the next-generation F-16 trainer. The system was prototyped using a big multipipe IR system (seven racks, just for the IG!), but it was to go into production on Bali.
Long story short, Rayt^W my former employer lost the contract to Lockheed, so all of those pre-orders for Onyx3s with Bali disappeared. I'm sure if that customer hadn't bailed, SGI wouldn't have been so quick to can the Bali project.
I agree with you that desktop graphics have come along way, but I think SGI still has the lead in terms of ability to render a quality visual scene in real time. Pure geometry performance is important, but since the poly count is fairly low in real-time sims anyway, texture and fill performance-- and high-speed texture cache size-- are even more important, and SGI wins those contests hands down.
Well, as I said before, I'm not a graphics programmer-- I'm a different kind of programmer-- so I may get some of these details wrong.
When you say "32 bits per pixel," you're talking about output pixel depth and format. A pixel in RGBA8 format stores one byte for each of red, green, blue, and alpha, and no other data. Those 32 bits are used by the DACs on the hardware to generate a component RGB video signal to drive your monitor. (Or, as I said before, a digital signal, but I'm not familiar with digital signal formats, so I get a little fuzzy at that point.)
IR doesn't support RGB8 or RGBA8; it uses either RGB10 (the default), in which 10 bits are used for each of red, green, and blue (not sure of the packing used), RGBA10 (adds alpha), or RGB12 (12 bpp).
On top of the color data, you can have a second buffer (used to eliminate image flicker in real-time animations), stereoscopic buffers (rendering two different images into the same buffer and display them through special stereo viewing hardware), auxiliary buffers (used for off-screen rendering in hardware; glCopyPixels() can copy aux buffer pixels into the visible frame buffer), multisample antialiasing, Z-buffering, and so on.
As I understand it from my vis sim buddies, it's really not that hard to fill up a 256 bit pixel in a real time image generator. They use 1 Kbit and 2 Kbit pixels pretty often.
Here's an example of a visual available on my Onyx2 at the office:
Visual ID: 6b depth=24 class=TrueColor
bufferSize=48 level=0 renderType=rgba doubleBuffer=1 stereo=1
rgba: redSize=12 greenSize=12 blueSize=12 alphaSize=12
auxBuffers=1 depthSize=23 stencilSize=8
accum: redSize=32 greenSize=32 blueSize=32 alphaSize=32
multiSample=4 multiSampleBuffers=1
Opaque.
I wish I could tell you what everything in there means, but most of it is beyond me.
Why is it sad that SGI designed and built graphics hardware in 1996 that is still technically superior to any graphics system on the market today? I don't think that's sad. I think that's astounding.
I think your definition of "mainstream" is flawed. The software is commercially available. You can buy it right now. I have a system set up in the office down the hall from mine, as a matter of fact. There are hundreds of Inferno systems around the world being used for motion picture and television production. Hell, the next season of Enterprise is being produced in HD, and (starting sometime in the next couple of weeks or so) effects will be created with Infernos at a post production house on Sunset Boulevard.
These products are absolutely, 100% mainstream. The fact that they're expensive doesn't make them not mainstream. The fact that there's no open-source-gimme-gimme version certainly doesn't make them not mainstream.
God created pointers to train the faithful, damnit.
LOL. That's now permanently inscribed in my fortune file.
Combine that with SGI's sudden idea to start selling x86 systems running linux, and I was sure that they were only a year or two away from closing shop.
Yeah, it's interesting that there's a distinct correlation between SGI's dropping all of their x86-based workstations and servers and their being able to actually squeak out a tiny profit one recent quarter.
I guess the lesson there is, "don't build stuff that people won't buy."
You got most of your info about IR4 right, but I just wanted to clarify some things in greater detail. IR is confusing at first, and very different from the typical single-board graphics systems that most people are familiar with. All the details can be found here, but here's an executive summary.
;-)
InfiniteReality (be it the original IR on Onyx, or IR2, IR2E, IR3, or now IR4) is comprised of a set of boards. In order to function, the set has to include one geometry engine (or GE), one raster manager (or RM), and one display generator (or DG). The GE board is where the graphics coprocessors live, and it's responsible for most of the 3D math. The DG converts the frame buffer into an analog RGB signal, or a CCIR-601 SD video signal, or, recently, a digital signal.
The RMs are the interesting part. The RM board holds both the frame buffer (80 MB on IR3, 2.5 GB on IR4) and the texture RAM (256 MB on IR3, 1 GB on IR4). A graphics pipe can include one, two, or four raster managers. When you add RMs, you increase frame buffer size (or the size of the raster you can render), but texture cache.
So a four RM graphics pipe will have 10 GB of frame buffer and 1 GB of texture cache, but that 1 GB of texture will be on each of the four RMs. So each texture you download will be stored, in parallel, on each of the four RMs. This keeps texture operations nice and peppy even when you're rendering into a 3840 x 2160 buffer. (That's four times more resolution than HDTV, if you're interested.)
Note, also, that these memories aren't combined. The TRAM and the frame buffer RAM are isolated in hardware. You can't store textures in the frame buffer, and you can't render in texture RAM. So saying that IR4 has a combined 11 GB of graphics RAM is not quite true, and slightly misleading. But only slightly.
The whole thing adds up to an incredibly flexible system. You can configure the graphics pipe as a relatively small raster of 2,048-bit-deep pixels, or an 8-million-pixel raster of 256-bit-deep pixels, or almost anything in between. You can render a truly giant image-- about 3K by 2K pixels, progressive scan, or even more than that if you're willing to live with interlacing-- with full antialiasing, multi-buffered. It's pretty.
(If all you want is pure geometry performance, for viewing giant CAD models and stuff in real time in a VR environment, SGI also has their InfinitePerformance line of graphics hardware for Onyx. But that's another topic.)
Okay, that's enough "Rah-rah, IR" for one night, with just one more little piece of trivia. InfiniteReality graphics has remained fundamentally unchanged since 1996 or so. The only exception is the change from an Everest bus host to an XIO host system. Every few years, SGI has increased the speed of the GEs, or the texture capacity on the RMs, or the performance of the DACs in the DG, but the system itself hasn't really changed at all in six or seven years. That's pretty amazing.
Spoken like a true guy-who's-never-read-the-Aqua-guidelines.
In order to be in line with the Aqua UI guidelines, you have to implement them all, completely. You can't just get kinda-sort close, throw a pinstripe background behind your toolbar and some gummy window decorations, and call it a day.
You should read the Aqua HI guidelines sometime. (Also available in PDF. They just might open your eyes.
BitGeek, I love what you wrote here. I've just finished posting a comment on this subject that I'm pretty pleased with. I suspect that you might agree with the ideas behind it. Would you mind giving it a look?