Unix was designed to operate over slow serial lines. That's why commands are silent when they work, by the way.
I wager that between the two, it would probably be MUCH faster for me to walk someone through installing drivers on Linux. I mean, I could send them to a website, or I could say "Type wget http://mydomain.com/nvidia.sh", which I have temporarily redirected to the right file. Hell, I can probably force it to skip the license.
It wouldn't still be around if it didn't meet people's needs.
Yes it would. And for what it's worth, Pre-OSX Mac isn't something I'd touch if I could help it.
Given a choice between doing an extremely rarely-needed task in 3 seconds with years of learning curve, or doing it in one minute in Excel, I'll pick Excel every time.
Don't think you learned to use Excel and GUIs without years of learning curve. You like Excel because you can leverage all that. Well, I know the CLI, and I can leverage it for all kinds of things. That one command may not be needed very often, but there are an endless number of things I need to do that I can accomplish with 3 seconds on the commandline, easily adding up to years of time when many of them would be five minutes to half an hour trying to make it work in Excel, then calling tech support or relatives to see if they can make it work in Excel...
it's trivial to run CLI commands in a GUI-- try running GUI tasks like photo editing or desktop publishing in a CLI!
To the extent that you mean, I have an answer for you:
$ gimp
Ok, I know I'm cheating here, but really, think about that. And it's certainly possible to create a command which launches the Gimp and its GUI without already being in a GUI.
It's trivial to run a CLI in a GUI. It's not trivial to run CLI commands in a GUI, unless you can show me a GUI way of combining 3-5 hyperspecialized programs to do one small task on a set of data.
GUIs currently lend themselves to large, monolithic programs, which have thought of 90% of what you want to do. CLIs have tons of tiny programs, each of which covers only a small part of what you want to do, but you can combine them to cover really anything you want to do, because at that point, you're programming.
However, before the flames start, I am not advocating a return to the CLI for absolutely everything. I do think it's still a useful skill, however, and when I have essentially no CLI (OS 9) or a CLI that sucks (Windows/NT/DOS), I feel as frustratingly restrained as I do when my X won't start and I'm restricted to a commandline only.
Every interface has its place, and there are at least a few of them that are absolutely irreplacable. Another one is the 3D interface -- all hype aside, you simply cannot do 3D modeling or mapping effectively with only a flat GUI, much less a commandline -- you really need to be able to fly around your creation.
If you're the type of person whose job consists mainly of combining and/or sorting strange random data files over and over again, you might have an argument.
It's not even my job, really. Consider things like this:
$ for i in */*/*.flac; do destdir=/ipod/`dirname \"$i\"`; mkdir -p "$destdir"; oggenc "$i" -o "$destdir/`basename \"$i\".flac"; done
I believe if I were to separate that out into a shell script, I could throw an ampersand at the end of that oggenc command to make this suddenly multithreaded.
Basically, the above command converts an entire music collection from Flac to Ogg Vorbis. I could easily do other conversions, and there's plenty more I could do here, but I think it serves to illustrate something. Even if I expand the above, it's maybe ten lines, and anyone who knows even a little bit about the commandline can begin to understand it and hack on it. But this kind of thing -- batch processing -- is something that Windows people have whole specialized applications for. The above would probably be some little freeware download, it'd be a megabyte or two, and it'd come with spyware.
With a proper commandline, it's a 10-line shell script.
I can't say I have to do that particular thing every day. However, there are things every day that are made significantly easier by Unix tricks under the hood, even if I eventually turn the one liner into a 50 line shell or Perl s
Installing a driver should not be considered "basic functionality" on any computer. If you have to go beyond the standard drivers, something's wrong, and you damn well better be able to handle at least a commandline+google, even on Windows -- or you shouldn't be touching the drivers.
And yes, in that respect, Linux installation is very broken. I'll remind you of the simple solution: Have it come pre-installed. Ultimately, Windows can be just as hard to install, but installing an OS is not something ordinary users should have to be subjected to.
By the way, this has nothing to do with GPLv3 -- and GPLv3 will not split the licensing any more than it already is. If diverse licensing is what's killing open source on the desktop, proprietary software should've been dead for 3 decades or so.
I'm on a custom-built nvidia desktop. Custom kernel: Reiser4, amd64, all kinds of tweaks and custom hacks I've accumulated over the year. Gentoo, mix of stable and unstable, all kinds of custom overlays (equivalent of Debian/Ubuntu repos) and again, custom hacks and shell scripts. I have a home directory that I've carried around for years.
Last night, I installed the latest nvidia drivers and Beryl. All I had to do was install the most recent nvidia drivers and set Beryl as my window manager.
However, I did have a problem. My previous nvidia drivers ran with my existing xorg.conf, no problems, full resolution (1600x1200). These ones ran at 1280x1024 until I told it to stop trying to autodetect my LCD's native resolution. I can't figure out if this is nVidia's fault or my monitor's fault, though it does seem bizarre and annoying that the nVidia binary blob has absorbed stuff relating to the monitor and PCI express.
So, if anything is broken, it's not Beryl/Compiz stability at all that's the issue, it's the drivers. There are things that break -- for instance, XvMC looks REALLY broken when you try to drag it as a wobbly window, but what do you expect? And I doubt Ubuntu has XvMC enabled by default for anything, and I doubt they could, at this point.
Column-oriented also lets you compress the data on the fly efficiently: all the records are the same data type (string, integer, date, whatever) and lists of same data types compress well, and uncompress typically far faster than you can pull them off disk, so you can just automatically do it for all the data and save both speed and time...
Given enough spare CPU cycles, yes. LZO compression is probably good for that. In fact, this is part of the theory behind which Hans Reiser claimed Reiser4 will be over twice as fast as any other filesystem -- the cryptocompress plugin.
It's worth mentioning that there are other ways of improving performance here than compression -- you could stick it on a RAID, for instance. Ultimately, it comes to a question of cost. Are the CPU cycles cheaper than the disk space and RAID controllers? (My guess is that with LZO and very compressible data, the answer is yes, absolutely.)
Updates are faster with row-oriented: you access the last record file and append something, or access an intermediate record file and update one "row" across.
This seems right. However:
Searches are faster with column-oriented: you access the record file for "Favorite slashdot admin" and look for entries which say "Phred", and then output the list of rows of data which match.
In row-oriented, you have a couple of ways of dealing with this. Indexes are one; a separate table is another. In fact, if you split each field out into a separate (id,value) table, you can essentially create a column-oriented database out of a row-oriented one.
To pull data out, you then have to look at all the column files and index in the right number of records, but that goes relatively quickly.
Exactly as quickly as in the above scenario. However, it could easily be slower than if you were to look up a record by an indexed column, then pull the whole column.
Indexes are useful, but column-oriented is more efficient in some ways. You don't have to maintain the indexes, and can just automatically search any column without having indexed it, in a reasonably efficient manner.
I call BS on this. Yes, it's faster to do a fulltext search on one column of names, but some sort of hashing/index system will improve the performance of a search in just about all circumstances. An example where this might not be the case would be, say, a serial number -- to look up something related to item n, just go to the nth row in that table -- but in this case, deletions mean you either need to leave holes in your database, or update every single serial number in that table (and everything referring to that number) -- which could easily involve searching the whole database, and would certainly be slower than an index.
The only advantage here is if you were doing a lot of updates on a very small column of short-ish data. It may get to the point where updates here are almost as fast as updates on an un-indexed row-oriented table would be, but looking something up by column would be much faster -- but still not as fast as indexed.
But that just seems like an incredible edge case, or at least, the kind of performance advantage you'd have to benchmark thoroughly to justify, considering what you're giving up.
And it seems to me like row-oriented with indexes is faster than column-oriented with indexes.
However, there is one advantage -- it's sort of automagically doing some normalization for you, or forcing normalization onto you. That, and the storage advantage, means this would be great for a general-purpose database -- a replacement for MySQL, say. But it looks to me like if you have a genius DBA, it would be possible to tweak even more performance out of row-oriented, unless column-oriented provides a way to emulate it to a point -- at which point, the distinction is really moot.
It seems to me like this is either bold-but-stupid marketing, or they have some other advantage than column-orientedness. One such advantage might be the grid-computing aspect -- at first glance, strictly column-oriented looks easier to distribute. But I really don't know.
Anyway... I do remember hearing that the scene was an accident. Basically, Harrison Ford had diarrhea at the time. It was actually supposed to be a nice long fight, swords vs Indy's whip, but when you gotta go...
Ah, the useless facts you pick up watching movies... After Morpheus' fully-VR PowerPoint-like talk about the Real World, and Neo gets unplugged and staggers around saying "I don't believe it..." Then Cypher goes "He's gonna pop" and Neo pukes... That was real. Apparently they had really bad chicken pot pies that day...
The index would be the shared, open source part, collaboratively crawled (via PC software or browser plugin) by everyone who elects to participate.
The real trick is making this truly open in the Freenet kind of way -- no centralized servers at all (other than existing DNS and such).
Think for a moment: Suppose Google allowed anyone to write a plugin of sorts to allow specialized kinds of searches, and extended their API to support any kind of frontend accessing these plugins. So, anyone could use Google's index in pretty much any way that the database is capable of (so long as they don't cause problems for anyone else).
Let's ignore technological issues of where Google's going to get the brute force to be able to handle everyone's pet algorithms, not to mention preventing trolls and their forkbombs from damaging the system, while still allowing enough flexibility for everyone else. Setting those aside, there's still one very big issue: Google controls the index.
And as long as it's on Google's servers, no one is really going to be satisfied that nothing fishy is going on. As much as I like Google as a company, that much is the truth, and is why I, personally, don't use my Gmail account for any email I really care about.
The real trick, then, is to have a situation where setting up your own private Google is like setting up your own website. When I setup a website -- email, web, IM, anything -- I just have to buy a domain, which basically means I'm registered with a registrar and with ICANN. However, all I have to trust them to do is point people to my DNS server, and after that, absolutely everything can be on my own servers, in my basement if I like.
I'm not sure I can even imagine how we could create a search -- a fast, usable search -- which is as distributed as that. Basically, it would have to be not much more vulnerable to manipulation and trolling than if I built my own Google-like index farm and stored absolutely everything locally. I sort of know how to do that kind of thing for files, for instance -- Freenet has theoretically solved that problem, but Freenet is slow as a dog, and I have no clue how to do anything like that for indexes. Even if you ignore the anonymity, how do you make it truly peer to peer, without the lag that implies?
In any case, I seriously doubt Defcon requires much in the way of performance. Like I said, I've basically done the opposite here -- I've run a mostly 2D, DirectX 9 game via Cedega, windowed, and if anything, it's faster under Beryl. So, it certainly could be reasonable for most games.
I expect that if we were to, say, run Quake 4 in a window, that's where we'd start to really see performance issues. (UT2004 does NOT run well in a window for me, despite being native Linux and OpenGL.)
However, Beryl has one thing that I hope Vista picks up on -- in recent SVN versions, I can disable indirection for a single window. Thus, I can actually tell it to not apply any effects to one window, but to just pass that window straight through, negating any performance hit for that game other than the fact that Beryl exists and is doing other stuff at the same time.
And how does running a compiler and doing your development as an administrator NOT make your system less secure?
Well, for one thing, how would someone exploit this situation?
Hole in the IDE or something. Fine. But that's about running additional software as an admin. It has nothing to do with whether that software is a compiler, an IDE, a debugger, or anything.
But that's irrelevant. I never said that running a compiler/IDE as admin is OK. I said that having a compiler/IDE, period, doesn't automatically make you insecure. I said this because you implied the opposite:
Running a development environment - is the LAST thing you want to do if you want your system to remain secure.
You did not say anything in that sentence about whether it's running as admin or not. You just said, effectively "Running a development environment makes you less secure."
There is sort of a rationale behind that, and there are Unix servers which avoid having a cc or gcc installed specifically because it's a security hazard. But I don't buy it. In a modern environment, you have compilers and scripting languages everywhere, and a web browser can be sufficient to be a "development environment". On Unix, if bash is at all usable, it's also more than enough for you to actually implement a web server -- and, by the way, apt-proxy used to be written in bash.
I suppose you could say that reducing unneeded stuff makes you secure -- and that by that logic, a webserver with gcc is less secure than one without gcc, simply because it's easier to make sure everything there is secure. However, that's a far cry from it being "the LAST thing you want to do", at least as far as security is concerned.
I'm curious if the same thing would happen on OS X... Also, what about DirectX windows? (Does DirectX even have a windowed mode?) I seem to remember all the uproar was about layering OpenGL on top of DirectX -- I'm guessing, looking at this now, that DirectX games would have to go through the same thing (if they run windowed).
From what I can remember benchmarking at before, my UT2004 is running at fullscreen, at decent speed, on Beryl on Linux. It does this because I told Beryl to not do any sort of indirection on fullscreen windows. (This took me from ~25 fps to ~50 fps, so I know it did something.)
(As far as I'm concerned, it should kill the indirection as soon as there aren't any effects on that window anymore. Of course this means a drop shadow would lag you... Maybe the indirection will get fast enough, eventually, that no one will care?)
I also have one old 2D MMO which I run in a window, and apparently the latest Beryl SVN has a keystroke for toggling indirection on a given window. Therefore, I should be able to easily run that game at full speed. However, it's old enough and slow enough that it really shouldn't matter, so I haven't checked out the SVN version. Also, there is a keystroke for taking an app fullscreen, which would presumably disable the indirection.
I mean, other than the "running a whole separate game" thing, there shouldn't be much performance degredation for simple GL apps -- no more, at least, than if you were actually running two windowed GL games side by side.
Also tried: xv works, with no noticeable performance hit until I started dragging the window around (dragging any window which does animation causes the whole WM to lag -- not significantly (down to maybe 10 FPS, which seems to be what OS X's UI operates at), but noticeable.
Also: Surprisingly, xvmc works, again with no noticeable performance hit. It does cause problems when dragging a window around, though -- I believe this used to work pretty much as expected, whereas in Beryl it's possible to figure out what's going on, but it's far from graceful.
Explorer is unique in the sense that when you authorize a file operation via UAC it doesn't elevate the entire explorer process. There are a bunch of reasons for this.
I'm sure there are. But there are also a bunch of reasons why what it does elevate should persist for a little while, especially when part of the exact same fucking operation.
For that matter, why not elevate a single window of Explorer?
Running a development environment - is the LAST thing you want to do if you want your system to remain secure.
Erm, WTF?
In what universe is it possible to have a secure system, but not a secure development environment?
Put more simply, just how do you think such a secure system was developed in the first place?
And just WHY should it be run as admin? Why the bloody hell do you need admin privileges to edit/compile/link?
See, this, I agree with. Certainly, it's not a good idea, security-wise, to run a development environment as admin. But under no circumstances should the availability of a compiler make your system less secure.
Let me put it this way: While it takes a bit of hacking, there are all kinds of things I can do, in Wine, under Linux, to make Windows programs more secure and unable to screw up each other or my Linux, and without giving the user a billion prompts.
Here's one really simple example: Cedega's "Point2Play" interface. Every game is installed in its own fake Windows installation, complete with a Program Files, WINDOWS dir, and so on. With symlinks and hardlinks, it should be possible to even share some things across installations, though I'm not sure if Cedega actually does this. And of course, while Cedega doesn't do it by default, it's certainly possible to disable all access outside that game's dir.
That would allow Microsoft to pretty much implement whatever system they wanted, while still providing options to run legacy programs. In fact, it should've been possible at any point in the past few years for MS to throw their weight around and create a Linux distro. After all, they aren't selling a kernel, and Wine already exists; certainly MS, who've been building compatibility modes into Windows for years now, could make Wine work flawlessly.
But instead of truly starting from the ground up, or stealing some of our better ideas, they instead extended their old system. Ok, fine, I'd probably do the same thing. But they seem to have not even bothered to test the thing.
Could they have made it work perfectly? Maybe not, but they could've done better than this. Frankly, this is an insult.
Oh, and for the record: Most Linux apps don't hardcode many absolute paths, and the ones they do can often be overriden by environment variables or commandline arguments. Thus, it's trivial to take a global installation and make it per-user, or for that matter per-group, or whatever else you want to do. Worst case, Linux has chroot; Windows doesn't.
Not even that -i is disabled by default, but that it's an option at all.
I'm sure there's a way to disable the Vista prompt, but I don't know what it is. In any case, Windows would tend to hide that deep in some configuration area -- I would hope to at least see a "don't ask me again" checkbox, but I doubt it.
Unix puts this kind of thing right there in your face. Let's suppose, for the sake of argument, that cp has -i on by default, and that there's a -y option to disable it (similar to/y on DOS/NT). In that case, the first time I see an "are you sure" box, I can run "cp --help" and see a quick summary of options I could use. From there, it's only a short step to aliasing "cp" to "cp -y".
This is one problem I have with most GUIs -- the preferences are kept nowhere near where they're actually used. The extreme example of this is about:config in Firefox.
Oh, and on the flip side, one amazingly GOOD thing Firefox does: It defaults to asking you once whether or not you want to do something that some users might object to -- for instance, submit a form over an insecure connection, or flip between http/https. However, it includes a checkbox that is something to the effect of "Ask me every time." If you allow the action, but ignore the checkbox, it will never ask you that question again.
That combines what I consider to be the great principles of UI: Always ask the user before doing anything unexpected (unless the user explicitly asked for it), and always find a way to present features to the power user (don't hide them away in about:config), but do sane things by default, and make it easy enough to use the default that average users aren't overwhelmed by advanced options, but advanced users don't have to hunt for them. Especially nice because sometimes power users want to be lazy.
It's simply embarrassing to be the one advocating computers and technology, and have to explain some absolutely retarded functionality, and all you can say is "I don't know why the fuck it does that, it just does."
What's worse, everyone is trying to emulate them.
I mean, I understand -- the PC hardware revolution was largely because MS could run on any "IBM compatible". But it does make you wonder what could've happened and despair -- for instance, what if Linux had been a few years earlier? What if Linux had taken the place of DOS/Windows as being the one familiar environment you could make work anywhere, so that we had the same situation in software as we do in hardware -- true vendor independence?
Couldn't really happen, I know, but I wish... Oh, how I wish...
In any case, that is why I will never officially support Windows. Windows bugs are just embarrassing, both that computer "professionals" are capable of such crap, and that it looks to the end-user like it's the fault of my software.
VGA is dying and out of 30 computer screens on display at major retailers only 2 or 3 are left that do VGA, cold hard facts.
First, I don't want to go back to VGA. I'm quite happy with my DVI, thanks. And DVI does not require HDCP -- I'll likely never see HDCP as long as I'm running Linux.
And second, I haven't looked since last year, but I suspect you're just plain wrong. My video card comes with two DVI ports, but it also comes with DVI->VGA connectors for them. My monitor comes with a DVI port, a VGA port, and a component video (RCA) port, as well as other inane things like speakers and a USB hub. For power, it uses the standard power cable that you plug into any computer.
As for fear mongering, would you kindly explain to me what the purpose of DRM can possibly be, other than to deliberately prevent me from doing certain things, so they can charge me for them? I mean, Sony can't possibly be afraid that I'm about to download 50 gigs worth of data over BitTorrent just to watch one two-hour movie, rather than renting it for a couple dollars -- and it would be even harder, and worse quality, if I ripped it through the so-called "analog hole". It seems infinitely more likely that they just want to ensure that I pay for the same movie twice -- once on Blu-Ray and once on UMD -- rather than rip the Blu-Ray, drop the quality a ton, and just copy it to a PSP (or a Video iPod)...
Cold, hard, facts: DRM sucks, and it is your fault. Yes, you, who are reading these words. Given the choice, none of us would want DRM on our media, so it is YOUR fault that it exists, because YOU are being lazy and taking a "meh, it'll work" attitude, rather than calling congressmen, the RIAA/MPAA, and your favorite artists and DEMANDING your media without DRM -- and DEMANDING the repeal of the DMCA.
Heard that SDL, OpenGL, OpenAL and rest of this open stuff is not directly accessing the hardware, but there's yet another set of layers in between.
I suppose that's possible, but it would be financial suicide -- see QuakeWars.
From what I've heard, fullscreen GL apps will be as fast as anything else, it's only windowed ones that will suffer. In any case, I don't think it's really up to MS -- ATI/Nvidia do the GL libs anyway, so they can always ship a working GL.
Then again, if they do kill GL performance, then I suppose I really do have a case for Linux. For one thing, Cedega for Direct3D doesn't seem to be such a huge performance hit, even if there is another layer there.
Aero glass is not a simple 2D windowing system like in XP, it's a fully 3D hardware accelerated system. Which means you're trying to run two games at the same time (sorta).
Shouldn't be. In fact, someone was commenting on how they were doing some sort of swapping in/out of the video card?
I want you, and everyone else here, to try a simple test: Download some of the first OpenGL tutorial demos -- you know, the ones which show you a triangle and a square -- and run those. Notice how your CPU usage is almost none? That's because it only has to render a single frame.
You can have all the glass in the world, but if it isn't changing, it shouldn't be using any resources at all beyond a flat 2D image.
Oh, and I've run multiple games at once, just to see how it worked. Surprisingly, it didn't slow me down too much. The only way I'd really expect a huge performance hit is if there was some transform happening on the game window itself -- for instance, if I drag it, and it wobbles, then that should definitely lag the game. But it should be reasonably fast once I drop it and it settles.
Of course you don't own the copyright to the movie when you buy a DVD.
I didn't say that. I said:
I doubt you actually own the copy to much of the stuff that you've bought.
By which I mean, it's entirely possible that you don't technically even own that one copy.
The copyright holder can give you a license to use the work under whatever terms they see fit.
That is true. However, they can also sell you a copy without giving up their copyright. This is how books work -- when was the last time you saw a license in a book? No, there's just a copyright notice. The only one who gets any kind of license is the publisher.
The closed licenses of software such as OS X, and also the GPL, are both examples of just that - licenses from the copyright holder to use the software under certain conditions.
Except in the case of the GPL, you're not required to accept the license in order to use the software -- only to redistribute it.
If you actually knew how to do a job, instead of whining at how useless your MSCE is -- your own preference does play a rather large role here. Is it better for your employer to have to fire you and hire a Windows admin, or have you spend months or years becoming as competent a Windows admin as you were a Linux admin? Or is it better for them to put your expertise to good use here?
Oh, and why is it that trolls always come out with this exact argument -- it's "religious" when people choose Linux (or BSD or OSX), but it's "practical" when they choose Windows? I bet if they were a Windows-only shop, Asking Slashdot for a Windows solution (since their product was migrating to Linux), you wouldn't see anyone call him "religious" or a "zealot".
Switching OSes means the users and the company have to adjust to a new OS instead of a new app. Would that be your choice if it'd been a Windows app moving to Linux? Think about it -- would you rather learn a new app or a new OS?
Also: Long-term, switching OSes means they have to pay for Vista (thousands of dollars for 400 users, unless they get a deal), and they have to keep paying the Microsoft tax for new versions. It seems very likely to me that if this establishment is around for very long, the sheer cost savings of sticking with the OS that's free forever will offset retraining -- even assuming they go for a proprietary app.
And realize, you're advocating ActiveX. Is that really what you want people to use? I mean, all other things equal, would you advocate IIS+ActiveX (oh, and hold everyone back to IE6, when IE7 is a "critical update") over Apache+standards?
Slashdot Mirror
What do you mean by "if Microsoft tried to apply exactly the same reasoning to Linux"?
Do you mean opening up Windows? Or burning Ubuntu CDs? Or what?
Unix was designed to operate over slow serial lines. That's why commands are silent when they work, by the way.
I wager that between the two, it would probably be MUCH faster for me to walk someone through installing drivers on Linux. I mean, I could send them to a website, or I could say "Type wget http://mydomain.com/nvidia.sh", which I have temporarily redirected to the right file. Hell, I can probably force it to skip the license.
Intel has the best supported video card on Linux... but nVidia's Linux support is currently WAY better than their Windows support, even for 64-bit.
Yes it would. And for what it's worth, Pre-OSX Mac isn't something I'd touch if I could help it.
Don't think you learned to use Excel and GUIs without years of learning curve. You like Excel because you can leverage all that. Well, I know the CLI, and I can leverage it for all kinds of things. That one command may not be needed very often, but there are an endless number of things I need to do that I can accomplish with 3 seconds on the commandline, easily adding up to years of time when many of them would be five minutes to half an hour trying to make it work in Excel, then calling tech support or relatives to see if they can make it work in Excel...
To the extent that you mean, I have an answer for you:
$ gimp
Ok, I know I'm cheating here, but really, think about that. And it's certainly possible to create a command which launches the Gimp and its GUI without already being in a GUI.
It's trivial to run a CLI in a GUI. It's not trivial to run CLI commands in a GUI, unless you can show me a GUI way of combining 3-5 hyperspecialized programs to do one small task on a set of data.
GUIs currently lend themselves to large, monolithic programs, which have thought of 90% of what you want to do. CLIs have tons of tiny programs, each of which covers only a small part of what you want to do, but you can combine them to cover really anything you want to do, because at that point, you're programming.
However, before the flames start, I am not advocating a return to the CLI for absolutely everything. I do think it's still a useful skill, however, and when I have essentially no CLI (OS 9) or a CLI that sucks (Windows/NT/DOS), I feel as frustratingly restrained as I do when my X won't start and I'm restricted to a commandline only.
Every interface has its place, and there are at least a few of them that are absolutely irreplacable. Another one is the 3D interface -- all hype aside, you simply cannot do 3D modeling or mapping effectively with only a flat GUI, much less a commandline -- you really need to be able to fly around your creation.
It's not even my job, really. Consider things like this:
$ for i in */*/*.flac; do destdir=/ipod/`dirname \"$i\"`; mkdir -p "$destdir"; oggenc "$i" -o "$destdir/`basename \"$i\" .flac"; done
I believe if I were to separate that out into a shell script, I could throw an ampersand at the end of that oggenc command to make this suddenly multithreaded.
Basically, the above command converts an entire music collection from Flac to Ogg Vorbis. I could easily do other conversions, and there's plenty more I could do here, but I think it serves to illustrate something. Even if I expand the above, it's maybe ten lines, and anyone who knows even a little bit about the commandline can begin to understand it and hack on it. But this kind of thing -- batch processing -- is something that Windows people have whole specialized applications for. The above would probably be some little freeware download, it'd be a megabyte or two, and it'd come with spyware.
With a proper commandline, it's a 10-line shell script.
I can't say I have to do that particular thing every day. However, there are things every day that are made significantly easier by Unix tricks under the hood, even if I eventually turn the one liner into a 50 line shell or Perl s
Installing a driver should not be considered "basic functionality" on any computer. If you have to go beyond the standard drivers, something's wrong, and you damn well better be able to handle at least a commandline+google, even on Windows -- or you shouldn't be touching the drivers.
And yes, in that respect, Linux installation is very broken. I'll remind you of the simple solution: Have it come pre-installed. Ultimately, Windows can be just as hard to install, but installing an OS is not something ordinary users should have to be subjected to.
By the way, this has nothing to do with GPLv3 -- and GPLv3 will not split the licensing any more than it already is. If diverse licensing is what's killing open source on the desktop, proprietary software should've been dead for 3 decades or so.
I'm on a custom-built nvidia desktop. Custom kernel: Reiser4, amd64, all kinds of tweaks and custom hacks I've accumulated over the year. Gentoo, mix of stable and unstable, all kinds of custom overlays (equivalent of Debian/Ubuntu repos) and again, custom hacks and shell scripts. I have a home directory that I've carried around for years.
Last night, I installed the latest nvidia drivers and Beryl. All I had to do was install the most recent nvidia drivers and set Beryl as my window manager.
However, I did have a problem. My previous nvidia drivers ran with my existing xorg.conf, no problems, full resolution (1600x1200). These ones ran at 1280x1024 until I told it to stop trying to autodetect my LCD's native resolution. I can't figure out if this is nVidia's fault or my monitor's fault, though it does seem bizarre and annoying that the nVidia binary blob has absorbed stuff relating to the monitor and PCI express.
So, if anything is broken, it's not Beryl/Compiz stability at all that's the issue, it's the drivers. There are things that break -- for instance, XvMC looks REALLY broken when you try to drag it as a wobbly window, but what do you expect? And I doubt Ubuntu has XvMC enabled by default for anything, and I doubt they could, at this point.
Given enough spare CPU cycles, yes. LZO compression is probably good for that. In fact, this is part of the theory behind which Hans Reiser claimed Reiser4 will be over twice as fast as any other filesystem -- the cryptocompress plugin.
It's worth mentioning that there are other ways of improving performance here than compression -- you could stick it on a RAID, for instance. Ultimately, it comes to a question of cost. Are the CPU cycles cheaper than the disk space and RAID controllers? (My guess is that with LZO and very compressible data, the answer is yes, absolutely.)
This seems right. However:
In row-oriented, you have a couple of ways of dealing with this. Indexes are one; a separate table is another. In fact, if you split each field out into a separate (id,value) table, you can essentially create a column-oriented database out of a row-oriented one.
Exactly as quickly as in the above scenario. However, it could easily be slower than if you were to look up a record by an indexed column, then pull the whole column.
I call BS on this. Yes, it's faster to do a fulltext search on one column of names, but some sort of hashing/index system will improve the performance of a search in just about all circumstances. An example where this might not be the case would be, say, a serial number -- to look up something related to item n, just go to the nth row in that table -- but in this case, deletions mean you either need to leave holes in your database, or update every single serial number in that table (and everything referring to that number) -- which could easily involve searching the whole database, and would certainly be slower than an index.
The only advantage here is if you were doing a lot of updates on a very small column of short-ish data. It may get to the point where updates here are almost as fast as updates on an un-indexed row-oriented table would be, but looking something up by column would be much faster -- but still not as fast as indexed.
But that just seems like an incredible edge case, or at least, the kind of performance advantage you'd have to benchmark thoroughly to justify, considering what you're giving up.
And it seems to me like row-oriented with indexes is faster than column-oriented with indexes.
However, there is one advantage -- it's sort of automagically doing some normalization for you, or forcing normalization onto you. That, and the storage advantage, means this would be great for a general-purpose database -- a replacement for MySQL, say. But it looks to me like if you have a genius DBA, it would be possible to tweak even more performance out of row-oriented, unless column-oriented provides a way to emulate it to a point -- at which point, the distinction is really moot.
It seems to me like this is either bold-but-stupid marketing, or they have some other advantage than column-orientedness. One such advantage might be the grid-computing aspect -- at first glance, strictly column-oriented looks easier to distribute. But I really don't know.
Was he actually a Samurai?
Anyway... I do remember hearing that the scene was an accident. Basically, Harrison Ford had diarrhea at the time. It was actually supposed to be a nice long fight, swords vs Indy's whip, but when you gotta go...
Ah, the useless facts you pick up watching movies... After Morpheus' fully-VR PowerPoint-like talk about the Real World, and Neo gets unplugged and staggers around saying "I don't believe it..." Then Cypher goes "He's gonna pop" and Neo pukes... That was real. Apparently they had really bad chicken pot pies that day...
The real trick is making this truly open in the Freenet kind of way -- no centralized servers at all (other than existing DNS and such).
Think for a moment: Suppose Google allowed anyone to write a plugin of sorts to allow specialized kinds of searches, and extended their API to support any kind of frontend accessing these plugins. So, anyone could use Google's index in pretty much any way that the database is capable of (so long as they don't cause problems for anyone else).
Let's ignore technological issues of where Google's going to get the brute force to be able to handle everyone's pet algorithms, not to mention preventing trolls and their forkbombs from damaging the system, while still allowing enough flexibility for everyone else. Setting those aside, there's still one very big issue: Google controls the index.
And as long as it's on Google's servers, no one is really going to be satisfied that nothing fishy is going on. As much as I like Google as a company, that much is the truth, and is why I, personally, don't use my Gmail account for any email I really care about.
The real trick, then, is to have a situation where setting up your own private Google is like setting up your own website. When I setup a website -- email, web, IM, anything -- I just have to buy a domain, which basically means I'm registered with a registrar and with ICANN. However, all I have to trust them to do is point people to my DNS server, and after that, absolutely everything can be on my own servers, in my basement if I like.
I'm not sure I can even imagine how we could create a search -- a fast, usable search -- which is as distributed as that. Basically, it would have to be not much more vulnerable to manipulation and trolling than if I built my own Google-like index farm and stored absolutely everything locally. I sort of know how to do that kind of thing for files, for instance -- Freenet has theoretically solved that problem, but Freenet is slow as a dog, and I have no clue how to do anything like that for indexes. Even if you ignore the anonymity, how do you make it truly peer to peer, without the lag that implies?
Are you on Vista?
In any case, I seriously doubt Defcon requires much in the way of performance. Like I said, I've basically done the opposite here -- I've run a mostly 2D, DirectX 9 game via Cedega, windowed, and if anything, it's faster under Beryl. So, it certainly could be reasonable for most games.
I expect that if we were to, say, run Quake 4 in a window, that's where we'd start to really see performance issues. (UT2004 does NOT run well in a window for me, despite being native Linux and OpenGL.)
However, Beryl has one thing that I hope Vista picks up on -- in recent SVN versions, I can disable indirection for a single window. Thus, I can actually tell it to not apply any effects to one window, but to just pass that window straight through, negating any performance hit for that game other than the fact that Beryl exists and is doing other stuff at the same time.
Thanks, but I'll wait till SP2 for an annoyance level of "Exactly what OS X and Linux has had for years, but we'll pretend we invented."
Well, for one thing, how would someone exploit this situation?
Hole in the IDE or something. Fine. But that's about running additional software as an admin. It has nothing to do with whether that software is a compiler, an IDE, a debugger, or anything.
But that's irrelevant. I never said that running a compiler/IDE as admin is OK. I said that having a compiler/IDE, period, doesn't automatically make you insecure. I said this because you implied the opposite:
You did not say anything in that sentence about whether it's running as admin or not. You just said, effectively "Running a development environment makes you less secure."
There is sort of a rationale behind that, and there are Unix servers which avoid having a cc or gcc installed specifically because it's a security hazard. But I don't buy it. In a modern environment, you have compilers and scripting languages everywhere, and a web browser can be sufficient to be a "development environment". On Unix, if bash is at all usable, it's also more than enough for you to actually implement a web server -- and, by the way, apt-proxy used to be written in bash.
I suppose you could say that reducing unneeded stuff makes you secure -- and that by that logic, a webserver with gcc is less secure than one without gcc, simply because it's easier to make sure everything there is secure. However, that's a far cry from it being "the LAST thing you want to do", at least as far as security is concerned.
I'm curious if the same thing would happen on OS X... Also, what about DirectX windows? (Does DirectX even have a windowed mode?) I seem to remember all the uproar was about layering OpenGL on top of DirectX -- I'm guessing, looking at this now, that DirectX games would have to go through the same thing (if they run windowed).
From what I can remember benchmarking at before, my UT2004 is running at fullscreen, at decent speed, on Beryl on Linux. It does this because I told Beryl to not do any sort of indirection on fullscreen windows. (This took me from ~25 fps to ~50 fps, so I know it did something.)
(As far as I'm concerned, it should kill the indirection as soon as there aren't any effects on that window anymore. Of course this means a drop shadow would lag you... Maybe the indirection will get fast enough, eventually, that no one will care?)
I also have one old 2D MMO which I run in a window, and apparently the latest Beryl SVN has a keystroke for toggling indirection on a given window. Therefore, I should be able to easily run that game at full speed. However, it's old enough and slow enough that it really shouldn't matter, so I haven't checked out the SVN version. Also, there is a keystroke for taking an app fullscreen, which would presumably disable the indirection.
I mean, other than the "running a whole separate game" thing, there shouldn't be much performance degredation for simple GL apps -- no more, at least, than if you were actually running two windowed GL games side by side.
Also tried: xv works, with no noticeable performance hit until I started dragging the window around (dragging any window which does animation causes the whole WM to lag -- not significantly (down to maybe 10 FPS, which seems to be what OS X's UI operates at), but noticeable.
Also: Surprisingly, xvmc works, again with no noticeable performance hit. It does cause problems when dragging a window around, though -- I believe this used to work pretty much as expected, whereas in Beryl it's possible to figure out what's going on, but it's far from graceful.
I'm sure there are. But there are also a bunch of reasons why what it does elevate should persist for a little while, especially when part of the exact same fucking operation.
For that matter, why not elevate a single window of Explorer?
Erm, WTF?
In what universe is it possible to have a secure system, but not a secure development environment?
Put more simply, just how do you think such a secure system was developed in the first place?
See, this, I agree with. Certainly, it's not a good idea, security-wise, to run a development environment as admin. But under no circumstances should the availability of a compiler make your system less secure.
As usual, really.
Let me put it this way: While it takes a bit of hacking, there are all kinds of things I can do, in Wine, under Linux, to make Windows programs more secure and unable to screw up each other or my Linux, and without giving the user a billion prompts.
Here's one really simple example: Cedega's "Point2Play" interface. Every game is installed in its own fake Windows installation, complete with a Program Files, WINDOWS dir, and so on. With symlinks and hardlinks, it should be possible to even share some things across installations, though I'm not sure if Cedega actually does this. And of course, while Cedega doesn't do it by default, it's certainly possible to disable all access outside that game's dir.
That would allow Microsoft to pretty much implement whatever system they wanted, while still providing options to run legacy programs. In fact, it should've been possible at any point in the past few years for MS to throw their weight around and create a Linux distro. After all, they aren't selling a kernel, and Wine already exists; certainly MS, who've been building compatibility modes into Windows for years now, could make Wine work flawlessly.
But instead of truly starting from the ground up, or stealing some of our better ideas, they instead extended their old system. Ok, fine, I'd probably do the same thing. But they seem to have not even bothered to test the thing.
Could they have made it work perfectly? Maybe not, but they could've done better than this. Frankly, this is an insult.
Oh, and for the record: Most Linux apps don't hardcode many absolute paths, and the ones they do can often be overriden by environment variables or commandline arguments. Thus, it's trivial to take a global installation and make it per-user, or for that matter per-group, or whatever else you want to do. Worst case, Linux has chroot; Windows doesn't.
Not even that -i is disabled by default, but that it's an option at all.
/y on DOS/NT). In that case, the first time I see an "are you sure" box, I can run "cp --help" and see a quick summary of options I could use. From there, it's only a short step to aliasing "cp" to "cp -y".
I'm sure there's a way to disable the Vista prompt, but I don't know what it is. In any case, Windows would tend to hide that deep in some configuration area -- I would hope to at least see a "don't ask me again" checkbox, but I doubt it.
Unix puts this kind of thing right there in your face. Let's suppose, for the sake of argument, that cp has -i on by default, and that there's a -y option to disable it (similar to
This is one problem I have with most GUIs -- the preferences are kept nowhere near where they're actually used. The extreme example of this is about:config in Firefox.
Oh, and on the flip side, one amazingly GOOD thing Firefox does: It defaults to asking you once whether or not you want to do something that some users might object to -- for instance, submit a form over an insecure connection, or flip between http/https. However, it includes a checkbox that is something to the effect of "Ask me every time." If you allow the action, but ignore the checkbox, it will never ask you that question again.
That combines what I consider to be the great principles of UI: Always ask the user before doing anything unexpected (unless the user explicitly asked for it), and always find a way to present features to the power user (don't hide them away in about:config), but do sane things by default, and make it easy enough to use the default that average users aren't overwhelmed by advanced options, but advanced users don't have to hunt for them. Especially nice because sometimes power users want to be lazy.
For this very reason.
It's simply embarrassing to be the one advocating computers and technology, and have to explain some absolutely retarded functionality, and all you can say is "I don't know why the fuck it does that, it just does."
What's worse, everyone is trying to emulate them.
I mean, I understand -- the PC hardware revolution was largely because MS could run on any "IBM compatible". But it does make you wonder what could've happened and despair -- for instance, what if Linux had been a few years earlier? What if Linux had taken the place of DOS/Windows as being the one familiar environment you could make work anywhere, so that we had the same situation in software as we do in hardware -- true vendor independence?
Couldn't really happen, I know, but I wish... Oh, how I wish...
In any case, that is why I will never officially support Windows. Windows bugs are just embarrassing, both that computer "professionals" are capable of such crap, and that it looks to the end-user like it's the fault of my software.
Does that actually apply to MSIs, or just EXEs? It's hard to tell from your link...
First, I don't want to go back to VGA. I'm quite happy with my DVI, thanks. And DVI does not require HDCP -- I'll likely never see HDCP as long as I'm running Linux.
And second, I haven't looked since last year, but I suspect you're just plain wrong. My video card comes with two DVI ports, but it also comes with DVI->VGA connectors for them. My monitor comes with a DVI port, a VGA port, and a component video (RCA) port, as well as other inane things like speakers and a USB hub. For power, it uses the standard power cable that you plug into any computer.
As for fear mongering, would you kindly explain to me what the purpose of DRM can possibly be, other than to deliberately prevent me from doing certain things, so they can charge me for them? I mean, Sony can't possibly be afraid that I'm about to download 50 gigs worth of data over BitTorrent just to watch one two-hour movie, rather than renting it for a couple dollars -- and it would be even harder, and worse quality, if I ripped it through the so-called "analog hole". It seems infinitely more likely that they just want to ensure that I pay for the same movie twice -- once on Blu-Ray and once on UMD -- rather than rip the Blu-Ray, drop the quality a ton, and just copy it to a PSP (or a Video iPod)...
Cold, hard, facts: DRM sucks, and it is your fault. Yes, you, who are reading these words. Given the choice, none of us would want DRM on our media, so it is YOUR fault that it exists, because YOU are being lazy and taking a "meh, it'll work" attitude, rather than calling congressmen, the RIAA/MPAA, and your favorite artists and DEMANDING your media without DRM -- and DEMANDING the repeal of the DMCA.
I suppose that's possible, but it would be financial suicide -- see QuakeWars.
From what I've heard, fullscreen GL apps will be as fast as anything else, it's only windowed ones that will suffer. In any case, I don't think it's really up to MS -- ATI/Nvidia do the GL libs anyway, so they can always ship a working GL.
Then again, if they do kill GL performance, then I suppose I really do have a case for Linux. For one thing, Cedega for Direct3D doesn't seem to be such a huge performance hit, even if there is another layer there.
Meh. Wake me up when Vista SP2 is out.
Shouldn't be. In fact, someone was commenting on how they were doing some sort of swapping in/out of the video card?
I want you, and everyone else here, to try a simple test: Download some of the first OpenGL tutorial demos -- you know, the ones which show you a triangle and a square -- and run those. Notice how your CPU usage is almost none? That's because it only has to render a single frame.
You can have all the glass in the world, but if it isn't changing, it shouldn't be using any resources at all beyond a flat 2D image.
Oh, and I've run multiple games at once, just to see how it worked. Surprisingly, it didn't slow me down too much. The only way I'd really expect a huge performance hit is if there was some transform happening on the game window itself -- for instance, if I drag it, and it wobbles, then that should definitely lag the game. But it should be reasonably fast once I drop it and it settles.
I didn't say that. I said:
By which I mean, it's entirely possible that you don't technically even own that one copy.
That is true. However, they can also sell you a copy without giving up their copyright. This is how books work -- when was the last time you saw a license in a book? No, there's just a copyright notice. The only one who gets any kind of license is the publisher.
Except in the case of the GPL, you're not required to accept the license in order to use the software -- only to redistribute it.
If you actually knew how to do a job, instead of whining at how useless your MSCE is -- your own preference does play a rather large role here. Is it better for your employer to have to fire you and hire a Windows admin, or have you spend months or years becoming as competent a Windows admin as you were a Linux admin? Or is it better for them to put your expertise to good use here?
Oh, and why is it that trolls always come out with this exact argument -- it's "religious" when people choose Linux (or BSD or OSX), but it's "practical" when they choose Windows? I bet if they were a Windows-only shop, Asking Slashdot for a Windows solution (since their product was migrating to Linux), you wouldn't see anyone call him "religious" or a "zealot".
Switching OSes means the users and the company have to adjust to a new OS instead of a new app. Would that be your choice if it'd been a Windows app moving to Linux? Think about it -- would you rather learn a new app or a new OS?
Also: Long-term, switching OSes means they have to pay for Vista (thousands of dollars for 400 users, unless they get a deal), and they have to keep paying the Microsoft tax for new versions. It seems very likely to me that if this establishment is around for very long, the sheer cost savings of sticking with the OS that's free forever will offset retraining -- even assuming they go for a proprietary app.
And realize, you're advocating ActiveX. Is that really what you want people to use? I mean, all other things equal, would you advocate IIS+ActiveX (oh, and hold everyone back to IE6, when IE7 is a "critical update") over Apache+standards?