Re: Since when UNIX is the "real" system that runs the "real" machines?

Why Unix sucks, my take:

• this news group ~ see the asterisk at the start of this sentence? see Singh's 'pixs' of his cats made with those ohsocute punctuation characters? The :-) and the ;-( and the.... oh god these people are so cute I could retch. The ultra conservative Unix sysadmin types have opposed, FOR DECADES, a communication standard based on 16 bits (or higher); newsgroups, by default and according to their Unix masters, may only work to the 8 bit (one stop bit) standard. Thus, no attachments. Thus UUE. Present, defend, shed blood, Code Warriors, Elitists, Aristocrats, you are defending outmoded, anachronistic standards, thus, not standards, but infantile egoism.

• While UNIX is the definition of snobbishness, to be spoken of in hushed terms, only by the inner circle of the many-ringed cult, those who possess multiple doctorates in computer science, Windows is a system of, by and for USERS, not EUNUCHS.

• The VI editor (don't swoon, now, gnurus) ~ need I say more! (I can you know, much more!)

• Pro/E trash directly inherited from UNIX: Pro/TABLE, an abomination from day one. While Unix fans have been prating, for years, about the superiourity of their operating system over DOS/Windows, the 80s (pre-Windows) gave us poor, illiterate, prols the use of shareware database and spreadsheet software that was ten times more professional than Pro/TABLE ever was or ever could be. That's because Unix users are all romantic gamesmen. They quickly develop a fancy for the quick and dirty convenience of some junk like Pro/TABLE, then get busy making up pretty stories to mythologize, memorialize and stabilize their sorry misdeeds. And, once again, because of infantile egoism and the fact that no Unix nerd can ever be wrong about anything (swords drawn, fight to the death), we live, for over a decade with the consequence. The Berline wall and Russian communism collapse; the Immortal Pro/TABLE survives. DEATH TO PRO/TABLE! JUNK! JUNK! JUNK! JUNK.......

CTRL-A, Activate window: MS Windows was never so dumb, so irretrievably simple that it couldn't figure out that the very last remaining window open was/is/nothing-else-but the 'active' window. Thank a decade of Unix/Unix programmers for this silly excrescence. And so stubbornly silly that you can do nothing else ~ not click in the window, not move it around, not access a menu ~ no, nothing but CTRL-A would wake that dead duck up. Yeah, duck you, Unix! Your fans (fanatics) brag like the monkey inheritors of the wonders of Ankhor Wat ~ they had nothing to do with it but it's THEIRS so everyone stay away. (Yes, your Highnesses, we do, we will, you are welcome to it. But, please Sirs, try not to bloody things up too much as you brawl over who invented it and who owns it.)

I could go on, it's a very long list, but I tire of this vituperative vitriolic. Sanjay, buddy, you wanted a, what did you call it? a flame war? think you got it, son.

David Janes, designer, San Diego CA

(more)

Ok, it appears that we've both done enough name-calling for the time being. Let's get to the actual issue at hand. The original poster did dismiss UNIX in a rather cavalier way. No doubt about that. He also pretty much admitted to not knowing much about UNIX, so his conclusion about UNIX being 'history', apart from being factually inaccurate, must have been pretty annoying to anybody who does know a few things about the issue.

I still don't exactly see how this justified attacking him personally instead of his several inaccurate assumptions, but that's just me, I guess. He gave the specs of his computer, which is always a good idea to do when you are discussing performance. I don't recall him basing his opinion of his machine's performance on its CPU MHz rating (which would, indeed, have been misleading).

The well-known fact that most of the modern-day engineering software originated on UNIX (or had pioneering predecessors that were UNIX-based) is not in and of itself an argument to continue to use it on UNIX. Yes, there was time when UNIX was pretty much the only choice for such software. However, the situation is no longer as clear-cut as it used to be. Every software company that wishes to stay in business is continuously reevaluating both its product and the OS it runs on. The OS's technological superiority is just one of many factors under consideration, and on balance UNIX is no longer as clear-cut a winner as it used to be. In point of fact, there are a number of applications for which UNIX is a clear-cut loser.

I mentioned command line interface on purpose. As a matter of fact, I still prefer to do some things in Windows from the command prompt. The reason I do it, though, is NOT because I believe that it is inherently superior to GUI, but because the modern-day GUIs (Windows, MAC, and various UNIX desktops) are still pretty restrictive when it comes to moderately advanced operations, or even to some fairly straightforward rules-based multiple file manipulation. It bothers me that so many computer professionals, especially UNIX people, fail to realize that it is not the command line that is so good: it is the modern-day GUIs that are still very much underdeveloped. The ONLY area where command line is truly long-term indispensable is low-level access required when due to some hardware or software malfunction only the most basic functions of the OS can be loaded (what is called 'Safe Mode' in Windows). The only reason command line is still used for more than that is that the very people who should be busily working on eliminating the need for it are instead inventing bogus pretexts for preserving it. My most hated phrase in the Microsoft's marketspeak is 'This behavior is by design'; unfortunately, UNIX developers are just as guilty (if not more so) of this high-handed and shortsighted attitude.

I am not aware of any current CAD systems that use command line as a main interface. I would appreciate it if you let me know which ones are you referring to. I have in the past worked with a couple of 2D CAD programs that did that and I find it hard to imagine how any such system can hold a candle ease-of-use- or productivity-wise to one with a well-designed GUI. With one proviso: I don't think current attempts by several CAD companies (with SolidWorks being the leader in this regard) to emulate MS Office GUI qualify as well-designed interface for CAD.

From your posts, Sanjay, it seems to me that you are first and foremost a computer pro and Pro/E is kind of a sideline for you. Please correct me if I have made a wrong assumption. You believe that the original poster has no idea whatsoever about the inner workings of Windows and how they compare to UNIX and you get angry because despite his ignorance in this regard he feels he can make a legitimate choice. Well, your assessment is probably true. I know it certainly applies to me. Not having any education worth mentioning in computer science, I certainly have no business discussing Windows and UNIX kernel architectures and their respective advantages and flaws. Does that mean I cannot make an informed choice of OS for my needs? In my opinion, no. All of us very frequently have to make buying decisions concerning equipment of which internal design we have no idea. The main criteria is: does it do what I need it to do reliably and with as little maintenance/repair effort on my part as possible? For most CAD tasks, a PC running Windows XP or 2000 fits this criteria quite well.

The makers of Pro/E themselves have said on many occasions that Pro/E (and CAD in general) is very much a linear process ill-suited to multithreading. As such, the multitasking advantages of UNIX have very limited value for it. By the way, this does not seem to be just PTC's opinion.

I understand why somebody running animation, rendering, FEA, or CFD apps all of which benefit a lot from multithreading on a multiprocessor platform would choose a UNIX workstation. What little rendering I dabble in gets done fast enough on my PC.

I built my current PC last March for approximately $1,600. Here are the specs: Pentium 4 3.06GHz, 1GB DDR-400 RAM, PNY Quadro4 980XGL video card,

Here is my typical day-to-day working environment : Since I mostly design custom automation machinery, my assemblies routinely have several hundred components, reasonably often - around 1,000 or slightly more. I can have such assembly open in one Pro/E window, its drawing - in the other, one or two subassemblies and a couple of parts (which accounts for 4-6 simultaneously open Pro/E windows), several bills of materials which I create as EXCEL spreadsheets (although I use OpenOffice instead of MS Office), my browser open on the Web page of a supplier of some component I am using in my design, Adobe Acrobat Viewer open with a manual for another purchased component, and sometimes AutoCAD with a 2D drawing of some other purchased component which I am modeling for my assembly.

On the other end of modeling spectrum, I've recently worked on a part which, when all was said and done, ended up 37MB in size.

I don't ever have to wait for my PC when I am switching from task to task (well, maybe the Web browser - my DSL connection could be faster).

Redrawing an assembly of this size in wireframe mode might take 10-15 seconds, and that is typically the longest time I have to wait for my computer (similar amount of time to regenerate said 37MB part). My display is set to 1280 x 1024 at 32-bit color and maximum refresh rate my (rather old) 21-inch monitor will accept. I have tried higher resolutions and didn't see any degradation in speed.

Since I installed it, XP hasn't crashed once. On several occasions my computer was running continuously for more than a week without any noticeable changes in the behavior of either Windows or Pro/E. Since I am the one paying my electricity bills I prefer to turn it off at night, so I don't know if longer intervals between reboots will create any problems. I've had 2 or 3 Pro/E crashes, which were annoying, but after the latest video driver / Pro/E build code update they seem to have gone away.

Can you give people like me a good reason to switch to OS we've never used before and to pay at least 3 times what I paid for my PC to get a system remotely comparable to mine productivity-wise? As a matter of fact, I think I am being overly charitable here: a UNIX workstation that will do what I have just described as fast or faster than my PC will probably run over $7,000 - $9,000.

By the way, I haven't paid $10,000 for Pro/E. I am using Pro/E Foundation II which was purchased from RAND's Engineering.com Web store for about $3,500 (taxes, shipping, handling and first year maintenance included). So the level of complexity I have just described is being handled (easily) without any advanced assembly tools. I am pretty sure that with Advanced Assembly module my PC will easily handle assemblies at least 3-4 times larger than the ones I've worked with so far.

Of course, for projects consisting of hundreds of thousands and millions of components, UNIX remains the only game in town. For now. However, I think you are underestimating the progress made by Windows. 10 years ago nobody would have imagined having this discussion: comparing to UNIX machines, PCs were a child's toy. 5 years ago some of the stuff typically done on UNIX machines became doable on PCs. Barely. I think the dynamics of the process speak for themselves. Your assumption that Windows will forever remain behind UNIX is, I am afraid, just that: an assumption. Regardless of the current Windows' deficiencies, from its current position on the market Microsoft can afford reworking entire Windows code much easier than most of the UNIX developers - a minor upgrade.

And finally. I have always wanted to find out how LINUX will compare to Windows. One of these days I probably will. However, I've already ran across articles written by people who tried comparing LINUX to Windows XP running typical office apps on the same hardware. Their finding was that the latest Red Hat (which Pro/E is supported on) boots and runs these apps slower than Windows XP. Of course, given Pro/E's UNIX heritage, it might actually run faster under LINUX. If I can find time for that, I'll definitely give it a shot, if only to run OCUS benchmark on it and see how a different OS will affect results on the same hardware.

-- Alex Shishkin

Yes, Unix was created by people who know CS and programming concepts. Don't you think an OS created by such people will be better than a knock-off of the MAC OS, that when you look under the hood is a beastly mutation of a 16-bit CP/M derivative? Smell the coffee, man.

Also, MAC OS itself has BSD Unix purring under its hood. Apple, the people who make the most user-friendly OS around, recognize the power of Unix.

Unix is the foundation for technical computing and networking that made the internet, pro/e, nastran, and scores of other codes possible.

Unix scales from desktop PC's through to supercomputers, and your code compiles and runs on these platforms. No problems, as long as the code was written properly.

The people that choose to stay with Windows are the eunuchs, cause they don't want to learn an operating system that enables the civilized world to do all the things that enable people to get things done.

vi is a text editor, not a word processor. But vi is used for basic text editing, right through to preparing text for typesetting. vi is a small program that is spectacularly efficient... if you spend the time to learn its command set. You can find, change, and manipulate text more powerfully than any comparable Windows-based editor. Any editor that runs on Windows that even competes is usually a port of an editor such as emacs which itself came from Unix.

I'm tiring of this nonsense too...

Don't blame Unix for PRO/E's design/usability flaws.

S.

It has the same nauseating bravado of the Microsoft executive who said, "Unix will be just another footnote in the history of computer science."

Maybe I was thinking of that when I flamed the original poster.

The original creators of Unix have since made new OS's such as Plan 9 and its successor, called Inferno ... real-time, distributed, and fits in less than 1 meg. Too bad its too alien to most people that it has no support from anyone.

Its all about pollution. Back in 1988, the net was a place primarily for academics, government and industry labs. Nowadays, the spam, disinformation, and general foolishness is the result of the proletariat having no sense of responsible use of the technology infrastructure.

To have not said something, is to abet people like that.

He was so rigorous, with "it runs like a champ"...

Whether the CPU goes at 300 Mhz or 3000 Mhz, the facts are that a desktop machine is spending 90% of its time waiting for memory. Upping the clock gains you tinier and tinier benefits in day-to-day real-world usage.

If, on the other hand, the system design allows better multitasking, and the OS offers much more stability, the user is overall more productive.

If you have a game-card mutated to pretending to be professional quality graphics card, you will suffer from onscreen artifacting which will affect the render quality. Unless the card uses floating point representations through and through, you will have inaccuracies on the display.

These are some of the factors I keep in mind, and why I prefer a Unix machine.

I am not citing the continued use of Unix for historical reasons. I am saying that Unix was developed on machines for technical development work, and to support multiple people using the machine for typesetting, software development, and batch processing of large jobs. That was the context in which Unix was developed. By being developed at Bell Labs to meet needs such as these, it automatically is more than able to meet the needs of the bourgeouis folk. Unix in the 70's could support 10 people on a machine with less power than a 286, all at the same time. Thats efficiency.

Try that with a Wintel solution. Watch tears come to the users eyes.

I understand the marketing reasons, but I will not buy into the religous type of argumentation that "90% of the world uses Windows" and therefore "millions of people can't be wrong" ... Pro/E is an instance of technical computing and power user type work. Windows is an inferior OS for this purpose.

Well HCI is something I only have one course in, and that was years ago. If you think that using a computer can be made as easy as on Star Trek, it probably can eventually, but it won't be running Windows.

Remember, OpenGL is from SGI, a RISC/Unix company in its heydey ...

All of the seminal advances that people associate with increases in visualization power or usability come from eclectic companies like Cray/SGI, Sun, or Apple. ... never, never is it Windows.

Microsoft's "But its a feature..." excuse is well known... after all didn't one version of their OS ship with thousands of bugs? Intel is shipping CPU's with 60-odd defects in the silicon design.

Bill Gates appears in the South Park full-length movie, and gets his reward for a "feature" in Windows...

I haven't personally experienced the high-handedness from the Unix camp. Take a look at all the open source Linux efforts out there. That should show the prevailing attitude in the Unix camp these days.

I was thinking of CATIA V4 ...

Correct... I think PRO/E Wildfire does a good job, but if it were me, I would be building in support for LCD shutter glasses for 3D, and voice activated commands as well. One CAD product out there for Windows does this.

You are remarkably, even uncannily correct. I dabble in many areas, but PRO/E is one I want to become very creative and skilled with, even though I am not a mechanical engineer by training.

Yes, in terms of leveraging your existing hardware investment, slapping PRO/E onto a Windows PC makes for a quick solution. But this line of reasoning does not speak to the future at all.

Digression:

But its perhaps also psychology. I am one of those who would rather learn SGML, even though XML is 80% as powerful, and much easier to learn. For me, the ideal of knowing the most powerful of the markup langauges is worth it.

End of digression.

A single user, working on a single part, on a single machine, yes, multi- threading does not give much back.

However, what if you are holding a respository of parts, each of which might be being analyzed with PRO/Mechanica, or rendered, or having its kinematics analyzed. With my Unix machine, I SHOULD BE ABLE TO render some complex assembly, or do an animation, while switching back to some other open part and continue to work on it.

But I am not able to do this with PRO/E 2001, which suggests that the software is not well designed in terms of threading. The best I can do with Unix is switch to some other totally different application and use it, while rendering is going on, since I can't even switch to another PRO/E window. Not particularly bright, this lack of threads in PRO/E, though Windows users will be none the wiser.

Who sets the agenda for what you do? The PC maker, Microsoft, or PTC?

Neither the application, the OS, nor the hardware should stand in the way of your design freedoms.

Not bad. But do not try to use it for anything other than a desktop machine. No ftp, no video conferencing or web collaboration. Maybe PRO/Mechanica would work, but if an analysis takes 4 hours to run, basically you will have to use another PC if you want to continue your work.

Say, you aren't running dual displays are you? I have PRO/E in one display, and the online help in another.

Also, be aware that a doubling of clock speeds gain you maybe 20% more speed over the slower CPU. You pay a premium for that extra clock.

Thats quite a lot of stuff open. The 1 GB of RAM is helping there. But don't confuse this with multitasking. You are largely context switching here, and that fast CPU is spending 90% of its time in the idle loop.

The nastiest thing I have done was design a wing section entirely using parametrics for the dimensioning, so its thickness could be easily varied for CFD analysis to determine lift/drag to find the best thickness for a given speed. 200 points over two wing sections, 2 equations per point,

As I say, if you were doing a kinematics or stress analysis, then the context switching costs would affect the responsiveness of the machine, and hence your productivity.

Windows routinely ships with dozens or hundreds of bugs. Microsoft wants the early adopters to be their beta-testers. Unix is at least 5 times more reliable than Windows. If you were working with 10 other designers, and were doing concurrent engineering, and the MS file server crashed, you might then see the skeletons in the closet that Windows has.

It does depend on 3 crashes over how long a time period.

I doubt you'd switch, but I would offer the following:

stability - load the machine up hard, do multiple things on it, have multiple users logged in, run a web server with VRML exports of your designs, conference with people with text, voice over IP, and video, watch training videos, ALL AT ONCE, with NO CRASHES, and NO SECURITY PROBLEMS.

scalability - working on a new car or aerospace system? Store all your parts in an Oracle database, run the biggest heaviest analyses with 64-bits of precision, use multiple 64-bit processors to do parallel simulations. Let your designs grow into the hundreds of gigabytes, if you want, no problem. Design your parts on the workstation, analyse them on servers, get your design totally prepped and ready for manufacturing sooner than your competition. Break the 4 GB barrier, process sizes on 64-bit hardware are much larger than any Wintel solution.

professional graphics - I turned up the quality settings on my Sun graphics card to the max, and noticed that the CPU was doing alot more work since no matter how much you zoom in on the part, the quality now stays high. Curves to not become polygonalized as you zoom in. Also, a Sun graphics guy told me about visualization problems when graphics cards do not do subpixel anti-aliasing. Sun's best cards do this, where one sample in a "sample buffer" is compared with 16 of its neighbours, and then anti-aliased to produce the very best quality image possible, before painting it to the screen. Apparently their XVR-4000 visualization system can render 3D scenes like in Toy Story in real-time ... its like Pixar's RenderMan software, IN HARDWARE ... this isn't just about CAD anymore, its about HoloDeck quality VR. Getting back to the discussion I do plan on cranking the settings on my PC, just to see how long it takes to do spins and rotations and zooms. The Ultra does slow down, but its still usable.

I will always remember the quote from some Microsoft exec about Unix becoming another footnote in the history of computer science.

Windows is a market driven product by a market driven company. They do not place good software design any where near the top of their priorities. Technical computing like CAD is something that can be done on Windows, but it won't do it particularly well. Windows was never designed to be portable to new hardware architectures. Unix and C were designed for exactly this.

Linux is probably Windows fiercest competition, and would be the best choice for running technical apps on 32-bit hardware.

Can it? I don't think so. Backwards compatability is the main factor that slows software advances down. They are just getting their 32-bit act together, as 64-bit hardware looms. Linux will change to 64-bit hardware much more readily than Windows will. It will cause some people pain, but since everything is open source, simply re-compile it, or grab binaries from somewhere on the net.

But the 64-bit market is going to be split between AMD's Opteron and Intel's Itanium ... more than a few RISC companies jumped onto Itanium, only to jump back off again, when it was both late and underpowered. Itanium is an attempt at a VLIW / post-RISC architecture, and requires a VERY SMART compiler to squeeze out competitive performance. AMD will remain backward compatible, but this will place upper limits on its long-term performance becuase of the extra control logic they have to put there for that backwards compatability.

And the market will take YEARS to build software that will run efficiently on the 64-bit native hardware, since it will be a tiny fraction of the market for a long time.

There are still the multi-processor and server-related design issues that Sun, IBM and HP know how to do, but MS has no clue about. With Sun servers, you can replace a bad CPU with the power on, and then continue where you left off.

Sun has been 64-bit since 1995... they slowly transitioned Solaris to

Remember, the Linux kernel is doing MUCH MORE than the Windows XP kernel is. So things like screen updates and application responsiveness might not be quite as fast, but thats perfectly normal. As with Solaris, things only become interesting when you start loading the machine down with tasks. Then the differences start to become more apparent.

If you can, please try to run two simultaneous instances of the benchmark on both OS's ... then the context switching will come into play as well.

S.