Haven't these absolute idiots in the media ever heard of kids playing Cowboys & Indians? Cops & Robbers? Since when have little boys not had toy guns growing up?
As someone who writes computer games, I find this extremely scary.
Maybe this is a little off topic, but it's a longstanding tradition here in the Free OS world to have religious wars about which one is better, traditionally based on speed, along the lines of -"Yeah, well Slashdot runs linux and serves up 500k pages a day dynamic content" -"Yeah, well cdrom.com runs FreeBSD and serves up 500 gigs a day of files, and on a machine with 1/4 the aggregate MIPS"
In this proud, fine, longstanding tradition, I would really like to see LinuxPPC whup the pants of MacOSX. It's based on Mach, right? Those microkernels were never that fast:)
Anyone care to do a nice, fair SMB/httpd/ftpd benchmark?
Perhaps MindCraft will do one for us:)
----------
Welcome to the Promised Land, MacOSX. I wonder if Apple really knew what they were getting into here...?
I'm one of the lucky few with both a CompUSA & a MicroCenter within 5 minutes of home (yay northern VA!). I've used both extensively.
MicroCenter is still pretty expensive (compared to the prices I usually pay for stuff) but it still kicks CompUSA's butt when it comes to Linux, and intelligence of sales staff.
I've walked into a CompUSA which has only an outdated RedHat box w/manual (pricy) and found that the employees know nothing: Q-"Hi, I'm looking for Linux, but without the manual. Just CD's in jewel cases" A-"You want CD jewel cases? Over there", indicating empty jewel cases for sale.
Then I went to MicroCenter. Q-"I'm looking for Linux" A-"Aisle 5" They had (last time I checked) RedHat, SuSE, the WalnutCreek BigOldDistro 4 CD set (yay!), TurboLinux and Caldera. I don't remember if they had a Debian CD from WalnutCreek or wherever, I think they did. I was very pleased:)
RH: simpler install (largely because you don't need to figure out dselect, and there are far fewer packages to choose from). Overall a good distro, but you're pushed to use their graphical configuration tools -- which aren't bad per se, but can be annoying to those used to hacking config files. But the graphical network configuration, system configuration, services, etc. tools can be a real timesaver to someone not familiar w/the/etc directory and unwilling to learn. It's possible to screw over redhat by messing up your rpm's badly, tho it's never happened to me -- you have to try pretty hard.
Debian: Approximately equivalent package-wise to RH with the rufus rpm repository also in the distro:). Many many many choices. While in RH install it's a good idea to go in detail and choose each individual package you want, in Debian it's a good idea to choose one of the preconfigured options. 3500 odd packages is it? Deb is way better than RPM, and can upgrade your system automatically, fixing all dependencies etc. as it goes. It's beautiful. Debian has no nice graphical config tools built in, but many of us prefer it that way.
Question: in apt-get a security hole? How does it guarantee it's not fetching evil packages? Is it only because you implicitly trust the servers you set in your conf file? I've wondered this for a while...
First, before I start bashing them, I have to say that any company that employs Alan, Rasterman, and all the rest is OK in my book. Period. They do their share not to just leach off the success of Linux, and we must respect them for that.
I started on Slack, and ran it for a long time, back when RedHat was way less important than it is now. I don't consider Slack particularly hard to install. I enjoyed running it. I didn't enjoy upgrading it in serious ways however. When comes to serious packages like libc's and whatever, I want to know I've killed old stuff before putting in new stuff, not go chasing around files that have moved. It becomes increasingly hard to keep a Slack system up to date as it ages.
I still run a Slack3.1 system at home. I must enjoy suffering.
I started using RedHat with 5.1. I disagree with the comment above basing 5.1; it always worked well for me. Granted, I'm sitting here at CMU and applying all the updates wasn't a big deal:). I've kind of developed an ideology for using RedHat: run the system. Don't let it run you.
I kill all rpm stuff on kernels (I do those the normal old fashioned way. just plain easier). I kill it mostly on X. I have a lot of video cards floating around as a game developer, and frequently swap them around. I can't deal with this XConfigurator @#$%, I use xf86config, and always found it perfectly straightforward. For that matter I skip the config proggy and just write out a config file (or hack an old one). They're not that bad. I can't wait around for RPM's of the latest XFree86 release to come out -- I needed the XSuSE servers as soon as they were available. For that matter after all this time I don't really quite understand the way RH's X works if you follow their setup:). Very little of my system resembles virgin RH5.1....
But it's nice to be able to select features from 5.2/whatever's fresh and just put'em in with a one liner, or easily switch versions of wine for whatever I need to work right.
I dislike immensely the stupid win95 lookalike default WM config. What a mistake. I know GNOME/KDE weren't available. I don't care. They shoulda chosen a clean, elegant fvwm2 config with a good menu and some nice wallpaper:). But this will improve in any case with one of two beautiful DE's installed by default.
The other thing I can't stand is the control-panel. Just gets in the way, more likely to whack out your scripts, makes scripts harder to read. I wanna read man pages and do them by hand dammit.
I've recently installed Debian on my second computer (can't do it to my main machine till I get a fresh HD, can't risk a non-working machine). I like it a lot. The package system is better. The distro is much easier to manage from config files. Installing a whole mess of packages is easier (I never got thru installing GNOME 1.0. never figured out quite which rpm's were essential, or quite what order to install them in. bleah. I think I eventually found the instructions but the GNOME people should put the essential stuff in a big tarball with easy rpm install intructions...end digression...). From a technological standpoint it's just a better distro IMHO.
What does this mean? I like RedHat. I've used their stuff a hell of a lot. But they're not invincible, and the better distro will always win.
This is Linux after all. Down deep, we use it 'cos it's better. Oh, and because we like really cool desktops:)
Sorry to be so rude, but I very much agree with the above. And I contend the Linux's FAT32 support blows Windoze's ext2 support out of the water.
To me, supporting a file system means being able to use it. As in, I have several FAT filesystems mounted right now. Works fine. Has worked fine for a while. Was definitely in RH5.1's kernel (2.0.36? 34? whatever), since I recall installing a friend's system w/it and mounting his FAT32 HD. But can Windoze use an ext2 disk? Or any of the other dozen fs's linux supports? No.
Now partitioning, that's a different issue. You gotta make room for an OS. Linux distro's (w/the recent exception of Caldera, bravo to them) don't generally include PartitionMagic, so you can't generally on-the-fly repartition you windoze machine to install linux. This doesn't surprise me. FAT32 was not at all meant to be dynamically resized. Perhaps there are some FS's that are, but I'm not familiar with any. The guys who wrote PartitionMagic figured out how to do it, and if you want it you can pay (or get Caldera:).
Can Windows repartition a computer that has only Linux installed to install? No. In fact, it'll probably screw up your MBR as a prize. So why should Linux be able to squeeze Windows out of the way?
You either use PartitionMagic, or leave room beforehand (as I did, and have since I first installed linux lo those many years ago) and you'll be fine.
I'm sorry some/.ers were rude to you, Jimmy. But you're really not acting like you're on top of the situation.
That Corel should make the WordPerfect suite entirely free for educational use?
I remember how excited I was, sitting in the Computer Systems Lab at Thomas Jefferson and hearing that Corel would release a free "personal" version of WP. Sure, it wouldn't have some of the features. But finally there would be a word processor for the masses in the CSL (which has been UNIX based since forever). And then I read the license agreement, and lo and behold, it's not free for educational use. What a bummer.
We should start an email campaign to Corel. Who's with me?
Yes, that many Intel boxes draw quite a bit and warm things up pretty good, but compared to what? An ARM would be a bad choice -- poor FP performance.
My (albeit limited) experience with supercomputers is that you just have to deal with the heat and power issues. My high school, Thomas Jefferson, has an old ETA10 (top of the line in 1988 or so, about as fast as a PII 450 now:). It's got its own frikkin' air conditioning unit in a special room!
They're building a replacement Beowulf cluster out of Celeron 450a's:)
Mr. Kane actually responded -- which I find amazing considering the volume of mail he likely got. Here is his response, and he makes some valid points:
Ari - Thanks for your comments. Unfortunately, keeping up with all the 'feedback' I receive is a full time job, so I will have to keep my comments brief.
Unfortunately, many people equate 'open souce' with free software. This is not a model that can last. Many companies supporting open source development are hardware companies that find it easy to get wider support for their products by getting on the Linux bandwagon. But can a company even the size of IBM that lost $1bn last year in their PC business afford this sort of development over the long haul? I doubt it.
Also, considering the many problems AMD has had over the past several quarters not making money, how long can they afford to provide a cheaper alternative? They may have captured the retail market (not a Linux stronghold I might add) but at what price? And when K7 comes out and is no longer compatible with the Intel socket/slot architecture, will they even be able to keep that?
I'm not against Linux or Open Source per se (I have Red Hat, NT and 98 all installed on my laptop) but the only reason why companies have been able to afford to sell things at a loss these days is because of the insanity of the stock market as the money is in the equity valuation, not in whether or not a company can be profitable. When that goes away, where's the money going to come from for software development?
I must disagree with you about the state of consumer choice and Open Source software. Open Source Software (OSS) does not represent a decision not to make money from selling software. Many companies do -- RedHat and Caldera in the US, Pacific HiTech in Japan, and SuSE in Germany are just a few.
Why does this work? It stems from a realization that the software market does not work in a traditional economic sense, nor anything remotely like the hardware market (an example of perfect competition if there ever was one). MicroSoft can sell as many copies as they want of Windows at essentially no cost, once it's developed. The box, CD and manual represent a negligable part of the $90 (much more for WinNT) cost of the software. The cost to them is in fact technical support -- which is why the technical support has gotten so bad recently, to the point where you must pay for every incident if you are a regular customer. This is what OSS Value Added Resellers actually sell. Anyone can download a copy of RedHat, but you have to pay if you want technical support. Even MicroSoft acknowledges this is a good idea -- their coming reorganization includes a whole division of "Knowledge Workers".
And what of the programmers who write free software? The argument that they won't because they'd rather be paid is invalid -- they already do. Linux runs on hardware from Personal Digital Assistants (the PalmPilot and Compaq's experimental Itsy) through destroying WindowsNT on desktops and servers (see ZD's own articles comparing NT and Linux as a Windows Networking server) through supercomputers among the 100 fastest machines in the world (IBM built a Linux supercomputer with off the shelf parts and a $40 RedHat CD that was as fast as a Cray during LinuxWorld Expo in San Jose a couple of weeks ago). The base of superior software already exists. Many programmers contribute the tools they need, written to solve their own personal requirements. Others donate their time for fun (such as my friend Ian Peters, a fellow Carnegie-Mellon University student and the GNOME Games package maintainer). Still others are employed by OSS VARs to increase the value of the product -- in this catagory are Alan Cox the Linux hardware guru, and a big chunk of the GNOME desktop environment team, all employed by RedHat.
I also note that Intel's perfect following of Moore's law, and the constant pricing of a "nice" computer system (used to be about $2500 here in the US) was an artifact of the way Intel made all the machines. Those chips cost Intel much less than they're selling them at. But for many years they had a monopoly and no pressure to cut prices. But along came AMD and Cyrix to cut into Intel's marketshare in the sub-$1000 value priced PC arena (the kind of machine that will soon make a PC a standard appliance in every home in North America and Europe), and all of a sudden there was competition. Intel has already lost the lead to AMD for market share -- and the rule of thumb about pricing is out the window.
As of now, processor technology can still be developed by the big guns of AMD, Intel and Cyrix on a Moore's Law track (works well, since it gives the engineers a target), but there are arenas in the hardware market where the law simply doesn't apply. In 3D hardware the product cycle is closer to 9 months, and each new product has many more than twice as many transistors, since the leading manufacturers such as 3Dfx, nVidia, Matrox and ATI are competing on a technological playing field with near constant pricing between them.
If you have yet to try Linux, I suggest you do. There really is something to Eric Raymond's "Cathedral and Baazar" model of software development. We don't use Linux because we're foolish lunatics. Millions of us use it because it's better. That's why Linux is gaining market share in corporate servers far faster than any other player.
Yours sincerely,
Ari Heitner ----------- DC: 703/5733512 CMU: 412/8623003 www.singularity-software.com ----------- "You know how your whole life flashes in front of your eyes before you die? That's just gdb unwinding the call stack . . . "
Compaq may have some corporate bulk. And their standard PCs make be nonstandard @#$% I wouldn't touch with a 10 foot pole.
I guess I don't like Compaq much at all.
But they do have a right to sell Wintel boxes, and their servers are all right.
And I do love Digital. Why are they a dinosaur? Because they make the fastest chips in the world? Because everyone's favorite HolyOS runs damn fast on the fastest chips in the world?
Ok, VMS is a continuing mistake. But they don't make me use it.
As long as Compaq supports Digital's WRL and cool toys like the Itsy, they're all right in my book.
Alpha is successful product line, representing the absolute bleeding edge in processor technology. There are quite a few industries out there which require that kind of flexibility, and Alpha has always been quite price competitive since Digital always sold individual chips to OEMs, and the machines used so much standard Intel hardware.
The Digital arm of Compaq has a hell of a commitment to Alpha, and a lot of R&D invested in the next gen 21364. Alpha is already supported by Linux and WinNT, and if Merced is a flop is well poised to become the 64bit successor to IA32, as companies demand 64bit power that works easily with their existing applications.
Win95/NT (as usually configured) just uses a more-or-less correct generic monitor setting. You still have to choose a specific one from the list for optimal settings. Of course Linux can just choose default standard monitor settings.
Video cards are a bigger problem in Linux. In most cases you have to know and tell it. This is sometimes more,sometimes less friendly. Perhaps installs should make a more frequent habit of starting X in VGA16 (as windoze does) to provide a friendlier interface for continuing setup.
Isn't the solution to this to check/proc/pci, see what the cards report, and match with the list of drivers to see what you think the user has:
Bus 1, device 0, function 0: VGA compatible controller: NVidia Unknown device (rev 4). Vendor id=10de. Device id=20. Medium devsel. Fast back-to-back capable. IRQ 10. Master Capable. Late ncy=64. Min Gnt=5.Max Lat=1. Non-prefetchable 32 bit memory at 0xe4000000 [0xe4000000]. Prefetchable 32 bit memory at 0xe6000000 [0xe6000008].
"Looks to me like you have a RivaTNT!"
Or do many installs do this these days? I seem to recall redhat will (tho it didn't know about my permidia2 when i installed it last). How about YaST? Of course many who use Debian or Slack have no problem telling it without it trying any risky assumptions.
I think a big part of the perception problem is just that corporate guys seem to be looking at slightly outdated linux distro's, which lack many video cards. Video card support has improved a great deal recently, especially w/X 3.3.3.1 (adding G200, Permidia2, TNT support, amond others). But even RH5.2 doesn't come w/X 3.3.3.1, so many people will find their installs lacking drivers for their vid cards. It's a shame, really, tho I know it's hard for RH to keep up with zillions of rapidly advancing OS projects:)
---
On a different note: USB. Works on the iMac, the one linux platform so far that really needs it. Under PC...well...we got a USB gamepad from Gravis (they gave us an armful of assorted pads/sticks at GDC). It promised "1 step install". We plugged it into a W95 machine. Didn't work. Plugged it into a 98 machine. Worked. No advantage over old-fashioned version. And I'm afraid support for those far-out USB peripherals will always lag, since no one who writes linux drivers owns them (too bad...digital speakers...cool...).
Result:
- Linux: Limited USB support where _absolutely_ necessary. Some issues w/USB code base (current one is a bit complex & hard to approach. Linus has tried to remedy matter some...)
- Win95 "with USB Support": anecdotally, USB doesn't work.
- Win98: USB works, completely extraneous (well, i could make a case I do _need_ those speakers:)
- WinNT: No clue. Anyone?
BTW, Alan Cox being Alan Cox, if you send him some cool USB speakers I bet he'll make'em work:) This goes for many other community members. I bet a lot of people would add peripheral support if the manufacturers would send them samples/specs. Has certainly worked in the past for other classes of peripherals (TV cards -- alan sez he just got a new one in his diary -- NICs, etc).
Scherer recently said to me "3Dfx should wake up and leave the wrapper guys alone. Nothing benafits 3Dfx more than having everybody running Glide games on slow, crufty wrappers for other boards. That way, 3Dfx's boards always appear way faster"
Makes sense to me.
...
I don't really think anyone will read this. Sometime I'll do an article on/. about it maybe....
WARNING: blatant plug coming.
Go check out our new linux demo w/sound and stuff.
from someone who's been dealing with 3D card manufacturers on a professional basis for some time now.
As I mentioned above, I've played with basically everything that's out now. I even hope to have a TNT2 and a G400 fairly soon (matter of weeks).
I'd like to dispel some of the myths/FUD (yes, I dare say it) that's been going around.
1. Yes, it's silly for card manufacturers to protect their hardware to the paranoid extent they do People end up reverse engineering it anyhow. Granted, 3D cards are pretty complicated beasts to reverse engineer, AFAIK no one's actually done it. But it does come under Fair Use, it's perfectly legal (I'm not a lawyer:). They should just release it and get over it.
2. I don't wanna here any more of this nonsense about 3Dfx not being 32 bit They have some crazy filtering in the rendering pipeline that makes what the Voodoo2 actually displays quite different from what you see in the framebuffer from your code. We never figured it out. We asked 3Dfx how it works. I don't know that they'll ever tell us:). But it's a fair tradeoff for fill rate: image quality on the Voodoo2/3 is very good, esp. w/multitexture (I can personally guarantee from experience on my personal computer right here in front of me that the Voodoo2 looks as good or better than the TNT).
3. Glide belongs to 3Dfx. The wrappers break the law. 3Dfx is perfectly within their rights That's just the way it is. If a game is written for Glide alone, it's 'cos the developers intended it to be that way.
4. The only reason game developers fail to support other API's is...well...there are some good reasons It's easy to call programmers lazy. And we support both Glide and D3D, so maybe the guys who only do Glide are lazy. But - Glide is way easier to write than D3D. Trust me, I just got back from a several hours of dealing w/DX6. Yuck. - Glide is faster anyhow (and was especially faster than DX5). - 3Dfx does a very convincing job of buying programmers' souls with toys and love. What can I say, they're cool guys. They've always taken care of us.
Alan Cox noted at one point recently that there were some major upgrades to Linux SMP (I think it happened somewhere in the middle of the 2.1.x series) that he said made Linux SMP _very_ nice for high numbers of CPUs, up to 16. I don't remember whether UltraLinux can boot w/more.
I think that Linus was talking about the hardware limitation of i386 architecture, and I'd say you're throwing your money away w/more than 2. But SPARC is _way_ better than intel for multiprocessing. Tho I dunno if even on an Ultra-450 16 processors is really all that worthwhile.
I'll give it a shot, and discuss some other things.
Cards 1st: 3Dfx Voodoo/Voodoo2/Voodoo3: Glide and D3D. Linux support via Glide. And there is a mini-(quake)-GL for playing GLquake, both under linux and win32. Voodoo2/3 support dual texturing. Voodoo3 is really just overclocked SLI'd Voodoo2's on the same board. We measure the exact same speed with a V3-2000 against 2 Voodoo2's, except that the V3 can do 1280x1024 and 1600x1200, which the SLI'd V2's can't. Note: these boards are so fast that the limitation on poly count is floating point power of the CPU. Those V3's can fill something like 60fps in 1600x1200. The only limitation is how many poly's you can set up with your CPU. My celeron450a doesn't come close to the limit on my SLI'd V2's. I'm not sure what I can say on 3Dfx's next generation board, Napalm, except that it'll be really cool. Now if Glide/Linux worked on the Alphas, that would be far out. Anyone ever try it? nVidia RivaTNT/TNT2:D3D, tho X servers exist (you can use them as 2D cards in Linux, as I'm doing now). The TNT is about as fast as a Voodoo2 running D3D. But Glide is twice as fast as D3D, so the Voodoo's still have an advantage. QuakeGL's exist for these as well as for the older Riva128 (about as fast as a Voodoo1). TNT2 has received favorable reviews, showing it's about exactly as fast as a Voodoo3 in D3D (at least the one I saw). There remains some debate. I hope to get a TNT2 in a few weeks and answer the questions for myself. BTW "TNT" stands for TwiNTexel, i.e. dual texturing. My understanding is that the TNT2 is a TNT on steroids (.25u process instead of.35, much overclocked, etc.). ATI Rage128: pretty fast, I'm waiting for mine to arrive so I don't know how fast. Is now standard issue for PowerMacs. D3D, QuakeGL, and a couple of Mac-specific API's no one cares about (what, me, biased?). Will continue to be a very important board, since ATI has key OEM deals. You can expect this board to be pretty much standard issue with low-end PC's, much as the old RagePro was. 3DLabs Permidia3: Unknown quantity. Will support D3D and presumably a full GL since 3DLabs really makes high-end GL boards. The Permidia2 was originally like the FireGL 1000 from Diamond or something. Matrox G200, G400: The G200 is a respectable D3D board, about 70% as fast as a Voodoo2 by our measurements. The G400 will rock, and have crazy bumpmapping features. We're trying to get a G400 before E3 (www.e3expo.com), and we have to promise Matrox that we'll write bump mapping support by E3:). Our main man Omar at Matrox tells us the G400 is 3x as fast as a G200. That would be pretty frikkin' fast. I'm not sure if there are mini/full-GL's for these boards. S3 Savage3D, S4: The Savage3D seemed like a decent board when we saw it at E3 last year, but we could only run about 5 frames of Fire and Darkness before it locked up. Alpha silicon, kinda buggy:). We never heard from S3 nor saw a Savage3D again. Now they're back with the S4. It's supposed to have bump I think, but I have never actually tested either board. D3D at least, I don't know what else. G200 Xserver is out there. Rendition V2200: Saw this at E3 last May. Supports D3D and RedLine, Rendition's own API. But they disappeared completely and we have heard nothing from them.
I think that's everything important.
Now: why Glide? And why do I think 3Dfx has a case here? Because 3Dfx didn't intend Glide to be an API for everyone else. Glide was their weapon against the bloat of D3D. And those developers who chose only to write for Glide made a decision about their software that these wrappers violate.
D3D is crufty. D3D is slow. The tiny subsets of GL supported to run Quake are not enough to do every game, and due to M$'s attempt to fsck GL on win32, they really link on top of D3D, so they're extra special slow.
Glide is not the answer, IMHO. GL is supposed to be the standard. Blasphemous as it is, To blatantly steal a line from Scherer, I really criticize id and Carmack for this one: id had the leadership position to demand that full, fast GL's be written for all major hardware and platforms. Instead, they chose to ask for a crufty, id-specific quasi-GL blob, useless to anyone except them.
No wonder developers choose to write for nice, clean Glide. It's easy, relatively friendly (or at least straightforward, since it's very low level), and literally 2x as fast as D3D.
But how much would it take for these companies to put a few coders on writing full GL's for win32 and linux? Look how much good it's done 3Dfx -- they're way out front as far as developer loyalty. And they've now got their own in-house guys doing Glide/Linux. In this case, trusting M$ when they say "oh don't worry, we'll take care of the API" is a really bad idea.
In freshman Bio in High School. It took a 1.5 hour block, no more.
You put the onion in the blender (yes, a kitchen blender) in a solution that breaks down the cell walls. Then you do some other stuff. Then you filter it, and you get white stringy stuff (_extremely_ fine threads) floating in solution at the bottom of a vial. I might have my vial sitting around at home somewhere.
It wasn't nearly as cool as the time Scherer & I failed miserably to clone African violets under a clean hood, and ended up growing a petri dish full of bacteria. Turned out we had moved our hands over the empty dish, and stuff falling off our hands (thoroughly scrubbed already BTW) contaminated the experiment.
It's the 10,500 rpm HD's that'll do ya in these days. I suggest mounting them in rails in a front accessable 5.25" slot, putting some static foam in front as a dust filter, and cooling the drives that way.
Two of my friends recently lost data to HD crashes because of heat. One was a Seagate Medallist Pro 7200rpm UDMA drive. The other was some 10krpm scsi drive. Both fine drives, but they run _very_ hot. Be careful.
Slashdot Mirror
The only journalistic voice of reason.
Even my home paper, the Washington Post, is guilty of poor journalism on this one.
We should definitely start an email campaign to authors of bad articles (and of good!) on this foolishness. Geeks have to protect our name.
Haven't these absolute idiots in the media ever heard of kids playing Cowboys & Indians? Cops & Robbers? Since when have little boys not had toy guns growing up?
As someone who writes computer games, I find this extremely scary.
Maybe this is a little off topic, but it's a longstanding tradition here in the Free OS world to have religious wars about which one is better, traditionally based on speed, along the lines of
:)
:)
-"Yeah, well Slashdot runs linux and serves up 500k pages a day dynamic content"
-"Yeah, well cdrom.com runs FreeBSD and serves up 500 gigs a day of files, and on a machine with 1/4 the aggregate MIPS"
In this proud, fine, longstanding tradition, I would really like to see LinuxPPC whup the pants of MacOSX. It's based on Mach, right? Those microkernels were never that fast
Anyone care to do a nice, fair SMB/httpd/ftpd benchmark?
Perhaps MindCraft will do one for us
----------
Welcome to the Promised Land, MacOSX. I wonder if Apple really knew what they were getting into here...?
I'm one of the lucky few with both a CompUSA & a MicroCenter within 5 minutes of home (yay northern VA!). I've used both extensively.
:)
MicroCenter is still pretty expensive (compared to the prices I usually pay for stuff) but it still kicks CompUSA's butt when it comes to Linux, and intelligence of sales staff.
I've walked into a CompUSA which has only an outdated RedHat box w/manual (pricy) and found that the employees know nothing:
Q-"Hi, I'm looking for Linux, but without the manual. Just CD's in jewel cases"
A-"You want CD jewel cases? Over there", indicating empty jewel cases for sale.
Then I went to MicroCenter.
Q-"I'm looking for Linux"
A-"Aisle 5"
They had (last time I checked) RedHat, SuSE, the WalnutCreek BigOldDistro 4 CD set (yay!), TurboLinux and Caldera. I don't remember if they had a Debian CD from WalnutCreek or wherever, I think they did. I was very pleased
I use both. Just my opinion tho...
/etc directory and unwilling to learn. It's possible to screw over redhat by messing up your rpm's badly, tho it's never happened to me -- you have to try pretty hard.
:). Many many many choices. While in RH install it's a good idea to go in detail and choose each individual package you want, in Debian it's a good idea to choose one of the preconfigured options. 3500 odd packages is it? Deb is way better than RPM, and can upgrade your system automatically, fixing all dependencies etc. as it goes. It's beautiful. Debian has no nice graphical config tools built in, but many of us prefer it that way.
RH: simpler install (largely because you don't need to figure out dselect, and there are far fewer packages to choose from). Overall a good distro, but you're pushed to use their graphical configuration tools -- which aren't bad per se, but can be annoying to those used to hacking config files. But the graphical network configuration, system configuration, services, etc. tools can be a real timesaver to someone not familiar w/the
Debian: Approximately equivalent package-wise to RH with the rufus rpm repository also in the distro
Question: in apt-get a security hole? How does it guarantee it's not fetching evil packages? Is it only because you implicitly trust the servers you set in your conf file? I've wondered this for a while...
what the point to running SCO is?
When is it a better choice than Linux, BSD, or Solaris?
How on earth do these guys still make money?
First, before I start bashing them, I have to say that any company that employs Alan, Rasterman, and all the rest is OK in my book. Period. They do their share not to just leach off the success of Linux, and we must respect them for that.
:). I've kind of developed an ideology for using RedHat: run the system. Don't let it run you.
:). Very little of my system resembles virgin RH5.1....
:). But this will improve in any case with one of two beautiful DE's installed by default.
:)
I started on Slack, and ran it for a long time, back when RedHat was way less important than it is now. I don't consider Slack particularly hard to install. I enjoyed running it. I didn't enjoy upgrading it in serious ways however. When comes to serious packages like libc's and whatever, I want to know I've killed old stuff before putting in new stuff, not go chasing around files that have moved. It becomes increasingly hard to keep a Slack system up to date as it ages.
I still run a Slack3.1 system at home. I must enjoy suffering.
I started using RedHat with 5.1. I disagree with the comment above basing 5.1; it always worked well for me. Granted, I'm sitting here at CMU and applying all the updates wasn't a big deal
I kill all rpm stuff on kernels (I do those the normal old fashioned way. just plain easier). I kill it mostly on X. I have a lot of video cards floating around as a game developer, and frequently swap them around. I can't deal with this XConfigurator @#$%, I use xf86config, and always found it perfectly straightforward. For that matter I skip the config proggy and just write out a config file (or hack an old one). They're not that bad. I can't wait around for RPM's of the latest XFree86 release to come out -- I needed the XSuSE servers as soon as they were available. For that matter after all this time I don't really quite understand the way RH's X works if you follow their setup
But it's nice to be able to select features from 5.2/whatever's fresh and just put'em in with a one liner, or easily switch versions of wine for whatever I need to work right.
I dislike immensely the stupid win95 lookalike default WM config. What a mistake. I know GNOME/KDE weren't available. I don't care. They shoulda chosen a clean, elegant fvwm2 config with a good menu and some nice wallpaper
The other thing I can't stand is the control-panel. Just gets in the way, more likely to whack out your scripts, makes scripts harder to read. I wanna read man pages and do them by hand dammit.
I've recently installed Debian on my second computer (can't do it to my main machine till I get a fresh HD, can't risk a non-working machine). I like it a lot. The package system is better. The distro is much easier to manage from config files. Installing a whole mess of packages is easier (I never got thru installing GNOME 1.0. never figured out quite which rpm's were essential, or quite what order to install them in. bleah. I think I eventually found the instructions but the GNOME people should put the essential stuff in a big tarball with easy rpm install intructions...end digression...). From a technological standpoint it's just a better distro IMHO.
What does this mean? I like RedHat. I've used their stuff a hell of a lot. But they're not invincible, and the better distro will always win.
This is Linux after all. Down deep, we use it 'cos it's better. Oh, and because we like really cool desktops
Sorry to be so rude, but I very much agree with the above. And I contend the Linux's FAT32 support blows Windoze's ext2 support out of the water.
:).
/.ers were rude to you, Jimmy. But you're really not acting like you're on top of the situation.
To me, supporting a file system means being able to use it. As in, I have several FAT filesystems mounted right now. Works fine. Has worked fine for a while. Was definitely in RH5.1's kernel (2.0.36? 34? whatever), since I recall installing a friend's system w/it and mounting his FAT32 HD. But can Windoze use an ext2 disk? Or any of the other dozen fs's linux supports? No.
Now partitioning, that's a different issue. You gotta make room for an OS. Linux distro's (w/the recent exception of Caldera, bravo to them) don't generally include PartitionMagic, so you can't generally on-the-fly repartition you windoze machine to install linux. This doesn't surprise me. FAT32 was not at all meant to be dynamically resized. Perhaps there are some FS's that are, but I'm not familiar with any. The guys who wrote PartitionMagic figured out how to do it, and if you want it you can pay (or get Caldera
Can Windows repartition a computer that has only Linux installed to install? No. In fact, it'll probably screw up your MBR as a prize. So why should Linux be able to squeeze Windows out of the way?
You either use PartitionMagic, or leave room beforehand (as I did, and have since I first installed linux lo those many years ago) and you'll be fine.
I'm sorry some
Any what's your bloody email address anyway?
That Corel should make the WordPerfect suite entirely free for educational use?
I remember how excited I was, sitting in the Computer Systems Lab at Thomas Jefferson and hearing that Corel would release a free "personal" version of WP. Sure, it wouldn't have some of the features. But finally there would be a word processor for the masses in the CSL (which has been UNIX based since forever). And then I read the license agreement, and lo and behold, it's not free for educational use. What a bummer.
We should start an email campaign to Corel. Who's with me?
Oh well, everyone there learnes LaTeX anyhow.
a mirror?
/. effect.
mmmmm,
Yes, that many Intel boxes draw quite a bit and warm things up pretty good, but compared to what? An ARM would be a bad choice -- poor FP performance.
:). It's got its own frikkin' air conditioning unit in a special room!
:)
My (albeit limited) experience with supercomputers is that you just have to deal with the heat and power issues. My high school, Thomas Jefferson, has an old ETA10 (top of the line in 1988 or so, about as fast as a PII 450 now
They're building a replacement Beowulf cluster out of Celeron 450a's
Mr. Kane actually responded -- which I find amazing considering the volume of mail he likely got. Here is his response, and he makes some valid points:
Ari -
Thanks for your comments. Unfortunately, keeping up with all the 'feedback'
I receive is a full time job, so I will have to keep my comments brief.
Unfortunately, many people equate 'open souce' with free software. This is
not a model that can last. Many companies supporting open source
development are hardware companies that find it easy to get wider support
for their products by getting on the Linux bandwagon. But can a company
even the size of IBM that lost $1bn last year in their PC business afford
this sort of development over the long haul? I doubt it.
Also, considering the many problems AMD has had over the past several
quarters not making money, how long can they afford to provide a cheaper
alternative? They may have captured the retail market (not a Linux
stronghold I might add) but at what price? And when K7 comes out and is no
longer compatible with the Intel socket/slot architecture, will they even
be able to keep that?
I'm not against Linux or Open Source per se (I have Red Hat, NT and 98 all
installed on my laptop) but the only reason why companies have been able to
afford to sell things at a loss these days is because of the insanity of
the stock market as the money is in the equity valuation, not in whether or
not a company can be profitable. When that goes away, where's the money
going to come from for software development?
-Bob-
Dear Mr. Kane:
I must disagree with you about the state of consumer choice and Open
Source software. Open Source Software (OSS) does not represent a decision
not to make money from selling software. Many companies do -- RedHat and
Caldera in the US, Pacific HiTech in Japan, and SuSE in Germany are just a
few.
Why does this work? It stems from a realization that the software market
does not work in a traditional economic sense, nor anything remotely like
the hardware market (an example of perfect competition if there ever was
one). MicroSoft can sell as many copies as they want of Windows at
essentially no cost, once it's developed. The box, CD and manual represent
a negligable part of the $90 (much more for WinNT) cost of the software.
The cost to them is in fact technical support -- which is why the
technical support has gotten so bad recently, to the point where you must
pay for every incident if you are a regular customer. This is what OSS
Value Added Resellers actually sell. Anyone can download a copy of RedHat,
but you have to pay if you want technical support. Even MicroSoft
acknowledges this is a good idea -- their coming reorganization includes a
whole division of "Knowledge Workers".
And what of the programmers who write free software? The argument that
they won't because they'd rather be paid is invalid -- they already do.
Linux runs on hardware from Personal Digital Assistants (the PalmPilot and
Compaq's experimental Itsy) through destroying WindowsNT on desktops and
servers (see ZD's own articles comparing NT and Linux as a Windows
Networking server) through supercomputers among the 100 fastest machines
in the world (IBM built a Linux supercomputer with off the shelf parts and
a $40 RedHat CD that was as fast as a Cray during LinuxWorld Expo in San
Jose a couple of weeks ago). The base of superior software already exists.
Many programmers contribute the tools they need, written to solve their
own personal requirements. Others donate their time for fun (such as my
friend Ian Peters, a fellow Carnegie-Mellon University student and the
GNOME Games package maintainer). Still others are employed by OSS VARs to
increase the value of the product -- in this catagory are Alan Cox the
Linux hardware guru, and a big chunk of the GNOME desktop environment
team, all employed by RedHat.
I also note that Intel's perfect following of Moore's law, and the
constant pricing of a "nice" computer system (used to be about $2500 here
in the US) was an artifact of the way Intel made all the machines. Those
chips cost Intel much less than they're selling them at. But for many
years they had a monopoly and no pressure to cut prices. But along came
AMD and Cyrix to cut into Intel's marketshare in the sub-$1000 value
priced PC arena (the kind of machine that will soon make a PC a standard
appliance in every home in North America and Europe), and all of a sudden
there was competition. Intel has already lost the lead to AMD for market
share -- and the rule of thumb about pricing is out the window.
As of now, processor technology can still be developed by the big guns
of AMD, Intel and Cyrix on a Moore's Law track (works well, since it gives
the engineers a target), but there are arenas in the hardware market where
the law simply doesn't apply. In 3D hardware the product cycle is closer
to 9 months, and each new product has many more than twice as many
transistors, since the leading manufacturers such as 3Dfx, nVidia, Matrox
and ATI are competing on a technological playing field with near constant
pricing between them.
If you have yet to try Linux, I suggest you do. There really is
something to Eric Raymond's "Cathedral and Baazar" model of software
development. We don't use Linux because we're foolish lunatics. Millions
of us use it because it's better. That's why Linux is gaining market share
in corporate servers far faster than any other player.
Yours sincerely,
Ari Heitner
-----------
DC: 703/5733512 CMU: 412/8623003
www.singularity-software.com
-----------
"You know how your whole life flashes in front of your eyes before you die?
That's just gdb unwinding the call stack . . . "
CC: Bob Kane of PCMagazine UK
Compaq may have some corporate bulk. And their standard PCs make be nonstandard @#$% I wouldn't touch with a 10 foot pole.
I guess I don't like Compaq much at all.
But they do have a right to sell Wintel boxes, and their servers are all right.
And I do love Digital. Why are they a dinosaur? Because they make the fastest chips in the world? Because everyone's favorite HolyOS runs damn fast on the fastest chips in the world?
Ok, VMS is a continuing mistake. But they don't make me use it.
As long as Compaq supports Digital's WRL and cool toys like the Itsy, they're all right in my book.
Alpha is successful product line, representing the absolute bleeding edge in processor technology. There are quite a few industries out there which require that kind of flexibility, and Alpha has always been quite price competitive since Digital always sold individual chips to OEMs, and the machines used so much standard Intel hardware.
The Digital arm of Compaq has a hell of a commitment to Alpha, and a lot of R&D invested in the next gen 21364. Alpha is already supported by Linux and WinNT, and if Merced is a flop is well poised to become the 64bit successor to IA32, as companies demand 64bit power that works easily with their existing applications.
I think it'll be around for a long time.
Win95/NT (as usually configured) just uses a more-or-less correct generic monitor setting. You still have to choose a specific one from the list for optimal settings. Of course Linux can just choose default standard monitor settings.
/proc/pci, see what the cards report, and match with the list of drivers to see what you think the user has:
:)
:)
:) This goes for many other community members. I bet a lot of people would add peripheral support if the manufacturers would send them samples/specs. Has certainly worked in the past for other classes of peripherals (TV cards -- alan sez he just got a new one in his diary -- NICs, etc).
Video cards are a bigger problem in Linux. In most cases you have to know and tell it. This is sometimes more,sometimes less friendly. Perhaps installs should make a more frequent habit of starting X in VGA16 (as windoze does) to provide a friendlier interface for continuing setup.
Isn't the solution to this to check
Bus 1, device 0, function 0:
VGA compatible controller: NVidia Unknown device (rev 4).
Vendor id=10de. Device id=20.
Medium devsel. Fast back-to-back capable. IRQ 10. Master Capable. Late
ncy=64. Min Gnt=5.Max Lat=1.
Non-prefetchable 32 bit memory at 0xe4000000 [0xe4000000].
Prefetchable 32 bit memory at 0xe6000000 [0xe6000008].
"Looks to me like you have a RivaTNT!"
Or do many installs do this these days? I seem to recall redhat will (tho it didn't know about my permidia2 when i installed it last). How about YaST? Of course many who use Debian or Slack have no problem telling it without it trying any risky assumptions.
I think a big part of the perception problem is just that corporate guys seem to be looking at slightly outdated linux distro's, which lack many video cards. Video card support has improved a great deal recently, especially w/X 3.3.3.1 (adding G200, Permidia2, TNT support, amond others). But even RH5.2 doesn't come w/X 3.3.3.1, so many people will find their installs lacking drivers for their vid cards. It's a shame, really, tho I know it's hard for RH to keep up with zillions of rapidly advancing OS projects
---
On a different note: USB. Works on the iMac, the one linux platform so far that really needs it. Under PC...well...we got a USB gamepad from Gravis (they gave us an armful of assorted pads/sticks at GDC). It promised "1 step install". We plugged it into a W95 machine. Didn't work. Plugged it into a 98 machine. Worked. No advantage over old-fashioned version. And I'm afraid support for those far-out USB peripherals will always lag, since no one who writes linux drivers owns them (too bad...digital speakers...cool...).
Result:
- Linux: Limited USB support where _absolutely_ necessary. Some issues w/USB code base (current one is a bit complex & hard to approach. Linus has tried to remedy matter some...)
- Win95 "with USB Support": anecdotally, USB doesn't work.
- Win98: USB works, completely extraneous (well, i could make a case I do _need_ those speakers
- WinNT: No clue. Anyone?
BTW, Alan Cox being Alan Cox, if you send him some cool USB speakers I bet he'll make'em work
all hail the holy gimp.
:)
Photoshop sux.
And is way out of my price range
Yay GIMP.
Who else loves the GIMP?
My last install of wine (and the dozen before that) have all taken 30 seconds:
:)
$rpm -Uvh wine-031599.rpm
I was very impressed when the last version I installed made MS Dev Studio start working
Scherer recently said to me "3Dfx should wake up and leave the wrapper guys alone. Nothing benafits 3Dfx more than having everybody running Glide games on slow, crufty wrappers for other boards. That way, 3Dfx's boards always appear way faster"
/. about it maybe....
Makes sense to me.
...
I don't really think anyone will read this. Sometime I'll do an article on
WARNING: blatant plug coming.
Go check out our new linux demo w/sound and stuff.
singularity-software.com.
from someone who's been dealing with 3D card manufacturers on a professional basis for some time now.
:). They should just release it and get over it.
:). But it's a fair tradeoff for fill rate: image quality on the Voodoo2/3 is very good, esp. w/multitexture (I can personally guarantee from experience on my personal computer right here in front of me that the Voodoo2 looks as good or better than the TNT).
As I mentioned above, I've played with basically everything that's out now. I even hope to have a TNT2 and a G400 fairly soon (matter of weeks).
I'd like to dispel some of the myths/FUD (yes, I dare say it) that's been going around.
1. Yes, it's silly for card manufacturers to protect their hardware to the paranoid extent they do
People end up reverse engineering it anyhow. Granted, 3D cards are pretty complicated beasts to reverse engineer, AFAIK no one's actually done it. But it does come under Fair Use, it's perfectly legal (I'm not a lawyer
2. I don't wanna here any more of this nonsense about 3Dfx not being 32 bit
They have some crazy filtering in the rendering pipeline that makes what the Voodoo2 actually displays quite different from what you see in the framebuffer from your code. We never figured it out. We asked 3Dfx how it works. I don't know that they'll ever tell us
3. Glide belongs to 3Dfx. The wrappers break the law. 3Dfx is perfectly within their rights
That's just the way it is. If a game is written for Glide alone, it's 'cos the developers intended it to be that way.
4. The only reason game developers fail to support other API's is...well...there are some good reasons
It's easy to call programmers lazy. And we support both Glide and D3D, so maybe the guys who only do Glide are lazy. But
- Glide is way easier to write than D3D. Trust me, I just got back from a several hours of dealing w/DX6. Yuck.
- Glide is faster anyhow (and was especially faster than DX5).
- 3Dfx does a very convincing job of buying programmers' souls with toys and love. What can I say, they're cool guys. They've always taken care of us.
Alan Cox noted at one point recently that there were some major upgrades to Linux SMP (I think it happened somewhere in the middle of the 2.1.x series) that he said made Linux SMP _very_ nice for high numbers of CPUs, up to 16. I don't remember whether UltraLinux can boot w/more.
I think that Linus was talking about the hardware limitation of i386 architecture, and I'd say you're throwing your money away w/more than 2. But SPARC is _way_ better than intel for multiprocessing. Tho I dunno if even on an Ultra-450 16 processors is really all that worthwhile.
I'll give it a shot, and discuss some other things.
.35, much overclocked, etc.). :). Our main man Omar at Matrox tells us the G400 is 3x as fast as a G200. That would be pretty frikkin' fast. I'm not sure if there are mini/full-GL's for these boards. :). We never heard from S3 nor saw a Savage3D again. Now they're back with the S4. It's supposed to have bump I think, but I have never actually tested either board. D3D at least, I don't know what else. G200 Xserver is out there.
Cards 1st:
3Dfx Voodoo/Voodoo2/Voodoo3: Glide and D3D. Linux support via Glide. And there is a mini-(quake)-GL for playing GLquake, both under linux and win32. Voodoo2/3 support dual texturing. Voodoo3 is really just overclocked SLI'd Voodoo2's on the same board. We measure the exact same speed with a V3-2000 against 2 Voodoo2's, except that the V3 can do 1280x1024 and 1600x1200, which the SLI'd V2's can't.
Note: these boards are so fast that the limitation on poly count is floating point power of the CPU. Those V3's can fill something like 60fps in 1600x1200. The only limitation is how many poly's you can set up with your CPU. My celeron450a doesn't come close to the limit on my SLI'd V2's. I'm not sure what I can say on 3Dfx's next generation board, Napalm, except that it'll be really cool. Now if Glide/Linux worked on the Alphas, that would be far out. Anyone ever try it?
nVidia RivaTNT/TNT2:D3D, tho X servers exist (you can use them as 2D cards in Linux, as I'm doing now). The TNT is about as fast as a Voodoo2 running D3D. But Glide is twice as fast as D3D, so the Voodoo's still have an advantage. QuakeGL's exist for these as well as for the older Riva128 (about as fast as a Voodoo1). TNT2 has received favorable reviews, showing it's about exactly as fast as a Voodoo3 in D3D (at least the one I saw). There remains some debate. I hope to get a TNT2 in a few weeks and answer the questions for myself. BTW "TNT" stands for TwiNTexel, i.e. dual texturing. My understanding is that the TNT2 is a TNT on steroids (.25u process instead of
ATI Rage128: pretty fast, I'm waiting for mine to arrive so I don't know how fast. Is now standard issue for PowerMacs. D3D, QuakeGL, and a couple of Mac-specific API's no one cares about (what, me, biased?). Will continue to be a very important board, since ATI has key OEM deals. You can expect this board to be pretty much standard issue with low-end PC's, much as the old RagePro was.
3DLabs Permidia3: Unknown quantity. Will support D3D and presumably a full GL since 3DLabs really makes high-end GL boards. The Permidia2 was originally like the FireGL 1000 from Diamond or something.
Matrox G200, G400: The G200 is a respectable D3D board, about 70% as fast as a Voodoo2 by our measurements. The G400 will rock, and have crazy bumpmapping features. We're trying to get a G400 before E3 (www.e3expo.com), and we have to promise Matrox that we'll write bump mapping support by E3
S3 Savage3D, S4: The Savage3D seemed like a decent board when we saw it at E3 last year, but we could only run about 5 frames of Fire and Darkness before it locked up. Alpha silicon, kinda buggy
Rendition V2200: Saw this at E3 last May. Supports D3D and RedLine, Rendition's own API. But they disappeared completely and we have heard nothing from them.
I think that's everything important.
Now: why Glide? And why do I think 3Dfx has a case here? Because 3Dfx didn't intend Glide to be an API for everyone else. Glide was their weapon against the bloat of D3D. And those developers who chose only to write for Glide made a decision about their software that these wrappers violate.
D3D is crufty. D3D is slow. The tiny subsets of GL supported to run Quake are not enough to do every game, and due to M$'s attempt to fsck GL on win32, they really link on top of D3D, so they're extra special slow.
Glide is not the answer, IMHO. GL is supposed to be the standard. Blasphemous as it is, To blatantly steal a line from Scherer, I really criticize id and Carmack for this one: id had the leadership position to demand that full, fast GL's be written for all major hardware and platforms. Instead, they chose to ask for a crufty, id-specific quasi-GL blob, useless to anyone except them.
No wonder developers choose to write for nice, clean Glide. It's easy, relatively friendly (or at least straightforward, since it's very low level), and literally 2x as fast as D3D.
But how much would it take for these companies to put a few coders on writing full GL's for win32 and linux? Look how much good it's done 3Dfx -- they're way out front as far as developer loyalty. And they've now got their own in-house guys doing Glide/Linux. In this case, trusting M$ when they say "oh don't worry, we'll take care of the API" is a really bad idea.
This "Anonymous Coward" is some sort of supergeek, posting near 24-hours a day!
He never stops!
Soon he will destroy us all!
Run for the hills!
In freshman Bio in High School. It took a 1.5 hour block, no more.
You put the onion in the blender (yes, a kitchen blender) in a solution that breaks down the cell walls. Then you do some other stuff. Then you filter it, and you get white stringy stuff (_extremely_ fine threads) floating in solution at the bottom of a vial. I might have my vial sitting around at home somewhere.
It wasn't nearly as cool as the time Scherer & I failed miserably to clone African violets under a clean hood, and ended up growing a petri dish full of bacteria. Turned out we had moved our hands over the empty dish, and stuff falling off our hands (thoroughly scrubbed already BTW) contaminated the experiment.
My Celeron 450a is quite reasonable.
It's the 10,500 rpm HD's that'll do ya in these days. I suggest mounting them in rails in a front accessable 5.25" slot, putting some static foam in front as a dust filter, and cooling the drives that way.
Two of my friends recently lost data to HD crashes because of heat. One was a Seagate Medallist Pro 7200rpm UDMA drive. The other was some 10krpm scsi drive. Both fine drives, but they run _very_ hot. Be careful.