Take a look at his posting history before you mod, and consider his truthfulness, verus the possibility that you're just feeding a troll, eh?
I'm also deeply impressed that people are sending email addresses to someone who four posts earlier claimed that he worked as a spammer.
Seriously, "a small business" that has a *legal department* at all, much less two lawyers, doesn't ring a bell here? Hell, I work at a rather large place that produces scads of patents, and we have *one* lawyer.
Nobody has had a problem with 8.3 for years. That's just a non-issue these days.
I've seen a *lot* of servers on Windows have security problems due to the fact that they'd check long filename paths but not equivalent 8.3 paths. I think it's safe to say that 8.3 is less of a glaring problem than it used to be, but it still causes plenty of damage.
I agree with you about the *NT* kernel ("win32 kernel" is nonsensical). I disagree that IIS doesn't have a lot of holes.
I *do* think MS is to blame WRT to Outlook viruses. The fact that the command to execute a program is the same as that to open a file is the root of the problem. Stripping executables is an annoying way to fix the thing. This is simply a security problem created by the fact that some system architect, in a simpler time when security was less of an issue, envisioned extending Explorer's reach to all throughout the OS for a competitive advantage. Having the same interface to open a file and run a program may be a decent idea in Explorer. It's significantly less good on a popup menu in an email program.
I'm not talking about add-on flaws. I'm talking about core problems with IIS and apache. Plus, apache is more resistant to add-on attacks as a whole, due to the fact that a UNIX system running apache can run the thing chrooted.
You know, I lived with an OpenBSD fan for some time, and one thing I never heard him trumpet OBSD's great advantages over Linux in was ease of use. I'm skeptical.
I disagree with MacOS WRT to trivial tasks. I'd say it starts to come into play around medium to power user level.
Huh...didn't know that the NT kernel understood symlinks -- I thought it was just hard links.
I'm not an NT systems coder, so I can't respond to some of these. I don't know why you'd need kernel-level support for thread pools. I don't agree that not having an OOM killer is more advanced design. I used to think that the OOM killer was an awful hack, but I've come to feel more and more that it's a lesser-evil solution. The overwhelming majority of software, and all large software packages that I can think of, simply do not check to ensure that they can succeed on all memory allocations. In a system where you don't have an OOM killer and run out of memory, things generally simply grind slower and slower, a couple apps get failed allocations (which may lead to crashes or odd behavior later in the lifetime of the app), and fairly soon something (generally one of the worse-written apps) crashes. So, essentially, you *have* an OOM killer on NT. It's just a bit less intelligent about choosing what to kill, and can wipe out more apps. The days of classic MacOS-style fancy application-level memory management are pretty much dead, IMHO. Too much programmer work. Finally, checking malloc() results doesn't do you a damn bit of good if the allocation is a stack allocation -- what are you planning to do, other than wedge the application or kill it?
Some of the things you mentioned seem to be features that are more useful in a Windows-like environment, where there's more of a focus on threads than processes. Given that UNIX coders have thread-based models and process-based-models available these days and tend to stick with process-based-models, I feel that this is more of a Windows flaw -- that if Windows allowed a decent fork(), process pools would eliminate the need for completion ports.
Finally, I think your argument supporting NT's file locking semantics is based on a misunderstanding of how UNIX file locking works. If I have a file open, it may be deleted. However, the file is refcounted, and each hard link and each open file table entry for that file counts as a reference. So the space for that file and all of its data remains valid until the application closes it. The only guarantee NT makes that UNIX doesn't is that if an application has a file open for read, then opening it again for read will not fail to deleted -- but it *could* change for a number of other reasons, like permission modifications. So NT's semantics provide very dubious benefits, and huge problems by way of forced rebooting and killing of applications.
Finally, WRT to FILE_SHARE_DELETE (which I admit that I did not know about) -- the thing is simply not a solution to UNIX semantics. First, it only works on NT, so no programmers will use it for anything but custom apps for at least a few years. Second, as far as I can tell, it requires the process deleting the file to take abnormal action to delete the file (OpenFile() with the FILE_FLAG_DELETE_ON_CLOSE). Third, I don't believe the file is deleted until it's actually closed (no Windows box handy to test on, however). This means that if you open the thing, you cannot create a new file with the same name in the same location until the process is closed. Fourth, it requires the process opening the file to take abnormal action (pass in the special flag). Fifth, there is a huge installed base of libraries and other functions that do not allow you to pass in FILE_SHARE_DELETE. The NT object loader doesn't do so, so DLLs cannot be replaced when an application is running (which means closing apps, rebooting computer, etc). It's a good bet that many userspace libraries also do not allow you to pass in this flag. This effectively makes it useless from the user's point of view.
My grandmother hasn't updated anything on her computer- she's 81 and more concerned with knitting and talking to her grandchildren. I just walked her thru an update.
Can you imagine if I had to tell her how to do that on linux ?? (without a subscription mind you)
I honestly feel that walking someone through typing the string "yum update" into a command prompt and hitting enter is much easier than walking them through Windows Update.
A not unreasonable chunk of the problem is not Windows itself -- it's that most technically inept users use Windows. If you dumbed down the Linux user base, you'd see at least some of the same problems.
You want QA on your kernels done by a QA team, you go to a distro vendor. The kernel was released by Linus, not by any vendors. That's the rough equivalent of doing a beta release.
Search for IIS on SecurityFocus's vulnerability database if you want a list of IIS holes. There are many.
The problem is that MS makes it very easy to do things that generally don't need to be done and have nasty potential effects (one-click running of executable attachments). Then then pop up warnings all over the place, and consider that a fix. When you delete a file. When you use Windows for the first time. When you disconnect a piece of hardware. When you view attachments. *I* don't read through them any more than I do the warning labels on every medicine bottle, potentially hot cup of McDonald's coffee, electrical product, etc. that I own, and I certainly don't expect the typical user to have to do so. The little warnings are no more than MS's cover-your-ass-legally.
I view the fact that DHCP failures on Windows are silent as a bad and nonintuitive thing, and can understand people getting irritated about it. I remember getting mildly irritated the first time I figured out what Windows was doing.
It'd be ideal if everyone could drop everything and IT could tell everyone to reboot all systems for a patch for a bug that needs to be deployed, but that just ain't generally the case.
And I'm saying that it's easy to accidently consider them part of the search. I did, and I was expecting something along those lines.
At some point, you have to say that being truthful but misleading goes too far. If you have something in your Policy Agreement twenty paragrahs in that says that the first seven hits are ads, I consider that not acceptable. Microsoft is numbering their ads as if they are actual search results. They label them as featured results. First of all, at least to my possibly not-cynical-enough brain, "featured" does not translate to "advertisement". It makes me think that perhaps this is a particularly useful or notable link. Second of all, when I search in Galeon, I see a medium-gray background with a hard-to-read, only slightly darker text stating that a given site is a featured site. It does not exactly stand out. (I cannot help but think that the fact that of all the websites I've seen in the past three months, the MSN one is the only one where using the default color choices in a non-IE browser is such that reading is annoying is somewhat nasty).
I agree that the author could have pointed out the "FEATURED" bit, but given that I made the same assumption, I don't think that it's all that egregious.
(even though GNOME has ridiculous problems like no wireframe mode, randomly-arranging taskbar, no way to disable the awful minimize animation, no main-menu keybindings etc.) when they're also funding an attempt at destroying the free software world.
Uh, oh. You seem to like KDE, and while I don't use GNOME as a DE, I generally feel that the GNOME application software is better than KDE.:-)
First of all, in GNOME 2, almost all power-user features are still present. You just need to edit a text file or run a command to get to them, instead of clicking a checkbox.
I'm not sure what WM you're referring to in not having a wireframe mode, but I assume it's not sawfish, which I use and know has one. If you're thinking of Metacity, this is in the GNOME FAQ:
7.2.3. To Reduce CPU Usage by Turning On Wireframe Mode
The Metacity window manager has a wireframe mode for when you move and resize windows. When wireframe mode is turned on, only the outline of windows is displayed when you move and resize windows. The contents of the window do not need to be updated during the move or resize. The contents of the window are displayed when the move or resize is complete.
To turn on wireframe mode, run the following command:
I'm not sure what problem you have with the taskbar -- I haven't used it for ages -- but it certainly wasn't randomly ordered when I used it.
The minimize animation toggle is similarly hidden away from non-power users:
Run 'gconf-editor' go to 'apps -> Metacity -> general' and change the 'animation_style' key to 'none'.
I'm not sure what you mean by main-menu keybindings. The GNOME menu? I haven't used the thing for ages, but I suspect there's a way to bind the thing to run apps when you hit a key. You might need to run gconf-editor and turn on desktop/gnome/interface/can_change_accels, highlight whatever you want to bind to a key in the menu, and then hit that key.
Open a terminal window in Windows to a directory. Then try to move or rename that directory (or a parent directory). An error will come up -- sharing violation. Linux will let you move the directory and simply use the new location.
Open a file, and try to move or rename that file while it's open (drives me nuts when using less in cygwin). Sharing violation. Doesn't happen in Linux.
Try running any kind of update or setup program. You generally get told to reboot. Why? Because Windows forces you to close all libraries, which means closing all programs using libraries, before they can be removed and the new versions of the libraries slapped into place. Linux uses UNIX file locking semantics, so the files can simply be deleted. They won't actually go away until the library is closed, but any new instances of programs started after an upgrade will use the new libraries. These poor file locking semantics are the reason for almost all of required Windows reboots.
I was particularly irritated when I noticed Microsoft's (IMHO dangerous and complexity-inducing) workaround for this. In XP, some MS exec realized that constant sharing violation error dialogs coming up when users tried to rename or delete files or directories containing open files or directories were pissing off users, so they ordered that this be fixed. Instead of fixing the NT kernel to be more capable, they made a workaround in XP's Explorer. From now on, failures in moving or deleting files and directories would be silent. Furthermore, to provide the user the illusion of his operation succeeding, XP's Explorer will even remove the directory's icon from any open windows. However, it is not actually deleted, and upon refreshing a window showing the directory's icon, you will notice that the icon returns.
I use about six Windows machines operating off of a single share on a regular basis. Since I frequently have consoles open in a directory (or Explorer windows open to a directory on other computers), I constantly get sharing violations. This is annoying and time-consuming, but harmless. However, Microsoft trying to play work around poor kernel design choices in Explorer is, IMHO, pretty awful.
I could also mention the poor workaround in Explorer for another NT kernel shortcoming -- the lack of support for symlinks (Shortcuts), and a host of other technical issues I have with Windows. (Remember the 8.3->long filenames issues?) The thing is that Microsoft isn't as interested in issues internal to Windows as Linus and friends do. Microsoft developers work to get a paycheck, and don't care what happens as long as end users don't see any obvious flaws. Linus is trying to produce a code showpiece, and if people can use it to help themselves out, fantastic. As a result, most Linux failings are due to the fact that developers weren't interested or motivated enough to deal with some issue that was of interest to end users but not developers, and most Microsoft issues are due to the fact that Microsoft made a customer-driven poor engineering decision in the past.
It's not -- it falls prey to the same issues of getting trapped that running simulated annealing without keeping things hot enough long enough runs into.
People *do* seek minima, but they will seek out local minima, not just global minima. If they're offered a feature that will make things easier and better with no cost to them, they'll take it. However, if they have the option to use something better but there is significant relearning time, they may well choose not to put out the effort.
In the case of Windows, many people know Windows. There are known issue with Windows where it does not fit with current best practices in human interface research. Take...oh, say, the use of pie menus, for instance. However, people are familiar with Windows's current linear menus, and even if there was a long-term benefit to changing to a different interface, they are going to be unhappy with the sort term cost.
I believe that the same thing is true of Linux.
How will Linux UI become "better" in the future? Why isn't these things implemented now?
The UI on Linux has been *steadily* (and compared to competitors, extremely rapidly) been improving. About twelve years ago, Linux didn't even exist. About ten years ago, you needed to be a bit of a kernel hacker to consider touching Linux. Seven years ago, a fairly serious techie experimenter, comfortable with poking around with your bootloaders. Five years ago, you had to still be a pretty decent power user, be comfortable not having a GUI for configuring much of anything, and be able to deal with lots of incompatibilities with Windows software, much less little hardware support. Four years ago, you had to be willing to deal with pretty alpha-ish, flaky or archaic desktop environment software, and still had to worry pretty constantly about hardware compatibility.
Frankly, Linux as a general user desktop environment has essentially gone from zero to threat #1 on Microsoft's worry list in the last three or four years. In some areas, UIs on Linux have surpassed their Microsoft equivalents. KDE's use of detachable panes or GNOME's complete user-configurability of keyboard menu equivalents are pretty neat. Four years ago, Linux multimedia was a pretty sad thing -- there was a commercial mpeg player called mtv and a couple of projects. Today, properly set up Linux boxes smoke Windows in latency. Microsoft has not adapted will to the tougher security requirements of an Internet-connected age, as Linux has.
Linux still has issues that keep some people from using it. A lack of entertainment software (most traditional video games do not make very good open source projects) is significant. Poor inter-distro binary and library compatibility is also an issue. If I had to ship something in binary format that I knew would run on Linux boxes, I'd probably ship it in PE format, because Wine can provide stronger guarantees about binary compatibility than Linux itself can. Linux also does not currently, IMHO, cater as well to the power-user-but-not-techie as Linux does. The light user, who uses a spreadsheet, word processor, email program, and web browser (oh, and Solitaire), has little problem with Linux other than an inability to interact with Microsoft Office file formats reliabily, and enjoys increased stability. The techie loves Linux's ability to be remotely administered, its performance, customizability, scriptability, huge (and free) suite of development tools, and availability of source to fix irritating bugs. The almost-techie-power-user, however, runs into problems. Linux has a thinner layer of GUI over the internals than does Windows. They're probably going to have to interact with the CLI. The power user may want to install unusual software, the sort of thing that doesn't come packaged, but be incapable of dealing with any problems in compiling that software.
So I believe that Linux is getting better for most desktop users much faster than Windows is, but there are definitely categories of users that will not be happy with Linux.
Yeah, Ikaruga is neat, but it's really a different sort of game from what I'm talking about. The traditional top-down shooter has huge, crazy, powerupable weapons (Ikaruga you have your special weapon, but that's it). Also, Ikaruga is a bit too hard for me. It's a really tough game.
Thanks for the links -- I knew about zinf, which is the old freeamp renamed (IIRC, development on it sort of foundered) The JuK website is certainly...brightly colored.
Mike, this looks very interesting indeed. If this is what I think it is, and it gets the support it needs, it will finally fix a longstanding and huge hole in Linux desktop usability.
Just so that I'm sure (I've skimmed a few Google searchs, and am still not sure), is this a fallback-capable system, or just a mixer with an ALSA interface? Does it use hardware channels as available until out of them and *then* do software mixing, or is it just another software mixer?
So. What's the damn use of having a PCI-X bus backwards compatible in therms of clock frequency if it's volt incompatible?
Probably due to the fact that it's pretty easy to change voltage, but a pain in the ass to generate a host-synchronized, extremely precise different clock signal. So it's easy to wrap an existing 5V system in a 3.3V card. It's a wee bit more difficult to take existing circuitry and redesign it to work at a different clock rate.
It doesn't affect users directly, just engineering firms.
Slashdot Mirror
Take a look at his posting history before you mod, and consider his truthfulness, verus the possibility that you're just feeding a troll, eh?
I'm also deeply impressed that people are sending email addresses to someone who four posts earlier claimed that he worked as a spammer.
Seriously, "a small business" that has a *legal department* at all, much less two lawyers, doesn't ring a bell here? Hell, I work at a rather large place that produces scads of patents, and we have *one* lawyer.
Nobody has had a problem with 8.3 for years. That's just a non-issue these days.
I've seen a *lot* of servers on Windows have security problems due to the fact that they'd check long filename paths but not equivalent 8.3 paths. I think it's safe to say that 8.3 is less of a glaring problem than it used to be, but it still causes plenty of damage.
I agree with you about the *NT* kernel ("win32 kernel" is nonsensical). I disagree that IIS doesn't have a lot of holes.
I *do* think MS is to blame WRT to Outlook viruses. The fact that the command to execute a program is the same as that to open a file is the root of the problem. Stripping executables is an annoying way to fix the thing. This is simply a security problem created by the fact that some system architect, in a simpler time when security was less of an issue, envisioned extending Explorer's reach to all throughout the OS for a competitive advantage. Having the same interface to open a file and run a program may be a decent idea in Explorer. It's significantly less good on a popup menu in an email program.
I'm not talking about add-on flaws. I'm talking about core problems with IIS and apache. Plus, apache is more resistant to add-on attacks as a whole, due to the fact that a UNIX system running apache can run the thing chrooted.
You know, I lived with an OpenBSD fan for some time, and one thing I never heard him trumpet OBSD's great advantages over Linux in was ease of use. I'm skeptical.
I disagree with MacOS WRT to trivial tasks. I'd say it starts to come into play around medium to power user level.
Huh...didn't know that the NT kernel understood symlinks -- I thought it was just hard links.
I'm not an NT systems coder, so I can't respond to some of these. I don't know why you'd need kernel-level support for thread pools. I don't agree that not having an OOM killer is more advanced design. I used to think that the OOM killer was an awful hack, but I've come to feel more and more that it's a lesser-evil solution. The overwhelming majority of software, and all large software packages that I can think of, simply do not check to ensure that they can succeed on all memory allocations. In a system where you don't have an OOM killer and run out of memory, things generally simply grind slower and slower, a couple apps get failed allocations (which may lead to crashes or odd behavior later in the lifetime of the app), and fairly soon something (generally one of the worse-written apps) crashes. So, essentially, you *have* an OOM killer on NT. It's just a bit less intelligent about choosing what to kill, and can wipe out more apps. The days of classic MacOS-style fancy application-level memory management are pretty much dead, IMHO. Too much programmer work. Finally, checking malloc() results doesn't do you a damn bit of good if the allocation is a stack allocation -- what are you planning to do, other than wedge the application or kill it?
Some of the things you mentioned seem to be features that are more useful in a Windows-like environment, where there's more of a focus on threads than processes. Given that UNIX coders have thread-based models and process-based-models available these days and tend to stick with process-based-models, I feel that this is more of a Windows flaw -- that if Windows allowed a decent fork(), process pools would eliminate the need for completion ports.
Finally, I think your argument supporting NT's file locking semantics is based on a misunderstanding of how UNIX file locking works. If I have a file open, it may be deleted. However, the file is refcounted, and each hard link and each open file table entry for that file counts as a reference. So the space for that file and all of its data remains valid until the application closes it. The only guarantee NT makes that UNIX doesn't is that if an application has a file open for read, then opening it again for read will not fail to deleted -- but it *could* change for a number of other reasons, like permission modifications. So NT's semantics provide very dubious benefits, and huge problems by way of forced rebooting and killing of applications.
Finally, WRT to FILE_SHARE_DELETE (which I admit that I did not know about) -- the thing is simply not a solution to UNIX semantics. First, it only works on NT, so no programmers will use it for anything but custom apps for at least a few years. Second, as far as I can tell, it requires the process deleting the file to take abnormal action to delete the file (OpenFile() with the FILE_FLAG_DELETE_ON_CLOSE). Third, I don't believe the file is deleted until it's actually closed (no Windows box handy to test on, however). This means that if you open the thing, you cannot create a new file with the same name in the same location until the process is closed. Fourth, it requires the process opening the file to take abnormal action (pass in the special flag). Fifth, there is a huge installed base of libraries and other functions that do not allow you to pass in FILE_SHARE_DELETE. The NT object loader doesn't do so, so DLLs cannot be replaced when an application is running (which means closing apps, rebooting computer, etc). It's a good bet that many userspace libraries also do not allow you to pass in this flag. This effectively makes it useless from the user's point of view.
Linux doesn't remotely compare to IIS. Apache would be a reasonable comparison, and is what I'm talking about.
Unfortunately, with Arpi gone, I'm not sure what's going to happen to mplayer development.
My grandmother hasn't updated anything on her computer- she's 81 and more concerned with knitting and talking to her grandchildren. I just walked her thru an update.
Can you imagine if I had to tell her how to do that on linux ?? (without a subscription mind you)
I honestly feel that walking someone through typing the string "yum update" into a command prompt and hitting enter is much easier than walking them through Windows Update.
He's right.
A not unreasonable chunk of the problem is not Windows itself -- it's that most technically inept users use Windows. If you dumbed down the Linux user base, you'd see at least some of the same problems.
Use a multi-clipboard program and you're fine.
You want QA on your kernels done by a QA team, you go to a distro vendor. The kernel was released by Linus, not by any vendors. That's the rough equivalent of doing a beta release.
Search for IIS on SecurityFocus's vulnerability database if you want a list of IIS holes. There are many.
I remember getting a CD with both the System 7.5.5 update from Apple and some Microsoft updates on the thing.
It's a shame that they don't mail updates to registered users, though. It'd be a good way to convince people to own legal copies.
The problem is that MS makes it very easy to do things that generally don't need to be done and have nasty potential effects (one-click running of executable attachments). Then then pop up warnings all over the place, and consider that a fix. When you delete a file. When you use Windows for the first time. When you disconnect a piece of hardware. When you view attachments. *I* don't read through them any more than I do the warning labels on every medicine bottle, potentially hot cup of McDonald's coffee, electrical product, etc. that I own, and I certainly don't expect the typical user to have to do so. The little warnings are no more than MS's cover-your-ass-legally.
I view the fact that DHCP failures on Windows are silent as a bad and nonintuitive thing, and can understand people getting irritated about it. I remember getting mildly irritated the first time I figured out what Windows was doing.
This was a two week old patch, mind you.
It'd be ideal if everyone could drop everything and IT could tell everyone to reboot all systems for a patch for a bug that needs to be deployed, but that just ain't generally the case.
Create a file named --help, then try to delete it.
touch -- --help
rm -- --help
And I'm saying that it's easy to accidently consider them part of the search. I did, and I was expecting something along those lines.
At some point, you have to say that being truthful but misleading goes too far. If you have something in your Policy Agreement twenty paragrahs in that says that the first seven hits are ads, I consider that not acceptable. Microsoft is numbering their ads as if they are actual search results. They label them as featured results. First of all, at least to my possibly not-cynical-enough brain, "featured" does not translate to "advertisement". It makes me think that perhaps this is a particularly useful or notable link. Second of all, when I search in Galeon, I see a medium-gray background with a hard-to-read, only slightly darker text stating that a given site is a featured site. It does not exactly stand out. (I cannot help but think that the fact that of all the websites I've seen in the past three months, the MSN one is the only one where using the default color choices in a non-IE browser is such that reading is annoying is somewhat nasty).
I agree that the author could have pointed out the "FEATURED" bit, but given that I made the same assumption, I don't think that it's all that egregious.
(even though GNOME has ridiculous problems like no wireframe mode, randomly-arranging taskbar, no way to disable the awful minimize animation, no main-menu keybindings etc.) when they're also funding an attempt at destroying the free software world.
:-)
/apps/metacity/general/wireframe_move_resize true
Uh, oh. You seem to like KDE, and while I don't use GNOME as a DE, I generally feel that the GNOME application software is better than KDE.
First of all, in GNOME 2, almost all power-user features are still present. You just need to edit a text file or run a command to get to them, instead of clicking a checkbox.
I'm not sure what WM you're referring to in not having a wireframe mode, but I assume it's not sawfish, which I use and know has one. If you're thinking of Metacity, this is in the GNOME FAQ:
7.2.3. To Reduce CPU Usage by Turning On Wireframe Mode
The Metacity window manager has a wireframe mode for when you move and resize windows. When wireframe mode is turned on, only the outline of windows is displayed when you move and resize windows. The contents of the window do not need to be updated during the move or resize. The contents of the window are displayed when the move or resize is complete.
To turn on wireframe mode, run the following command:
# gconftool-2 --type bool --set
I'm not sure what problem you have with the taskbar -- I haven't used it for ages -- but it certainly wasn't randomly ordered when I used it.
The minimize animation toggle is similarly hidden away from non-power users:
Run 'gconf-editor' go to 'apps -> Metacity -> general' and change the
'animation_style' key to 'none'.
I'm not sure what you mean by main-menu keybindings. The GNOME menu? I haven't used the thing for ages, but I suspect there's a way to bind the thing to run apps when you hit a key. You might need to run gconf-editor and turn on desktop/gnome/interface/can_change_accels, highlight whatever you want to bind to a key in the menu, and then hit that key.
Open a terminal window in Windows to a directory. Then try to move or rename that directory (or a parent directory). An error will come up -- sharing violation. Linux will let you move the directory and simply use the new location.
Open a file, and try to move or rename that file while it's open (drives me nuts when using less in cygwin). Sharing violation. Doesn't happen in Linux.
Try running any kind of update or setup program. You generally get told to reboot. Why? Because Windows forces you to close all libraries, which means closing all programs using libraries, before they can be removed and the new versions of the libraries slapped into place. Linux uses UNIX file locking semantics, so the files can simply be deleted. They won't actually go away until the library is closed, but any new instances of programs started after an upgrade will use the new libraries. These poor file locking semantics are the reason for almost all of required Windows reboots.
I was particularly irritated when I noticed Microsoft's (IMHO dangerous and complexity-inducing) workaround for this. In XP, some MS exec realized that constant sharing violation error dialogs coming up when users tried to rename or delete files or directories containing open files or directories were pissing off users, so they ordered that this be fixed. Instead of fixing the NT kernel to be more capable, they made a workaround in XP's Explorer. From now on, failures in moving or deleting files and directories would be silent. Furthermore, to provide the user the illusion of his operation succeeding, XP's Explorer will even remove the directory's icon from any open windows. However, it is not actually deleted, and upon refreshing a window showing the directory's icon, you will notice that the icon returns.
I use about six Windows machines operating off of a single share on a regular basis. Since I frequently have consoles open in a directory (or Explorer windows open to a directory on other computers), I constantly get sharing violations. This is annoying and time-consuming, but harmless. However, Microsoft trying to play work around poor kernel design choices in Explorer is, IMHO, pretty awful.
I could also mention the poor workaround in Explorer for another NT kernel shortcoming -- the lack of support for symlinks (Shortcuts), and a host of other technical issues I have with Windows. (Remember the 8.3->long filenames issues?) The thing is that Microsoft isn't as interested in issues internal to Windows as Linus and friends do. Microsoft developers work to get a paycheck, and don't care what happens as long as end users don't see any obvious flaws. Linus is trying to produce a code showpiece, and if people can use it to help themselves out, fantastic. As a result, most Linux failings are due to the fact that developers weren't interested or motivated enough to deal with some issue that was of interest to end users but not developers, and most Microsoft issues are due to the fact that Microsoft made a customer-driven poor engineering decision in the past.
That sounds like the most "optinal" UI to me.
It's not -- it falls prey to the same issues of getting trapped that running simulated annealing without keeping things hot enough long enough runs into.
People *do* seek minima, but they will seek out local minima, not just global minima. If they're offered a feature that will make things easier and better with no cost to them, they'll take it. However, if they have the option to use something better but there is significant relearning time, they may well choose not to put out the effort.
In the case of Windows, many people know Windows. There are known issue with Windows where it does not fit with current best practices in human interface research. Take...oh, say, the use of pie menus, for instance. However, people are familiar with Windows's current linear menus, and even if there was a long-term benefit to changing to a different interface, they are going to be unhappy with the sort term cost.
I believe that the same thing is true of Linux.
How will Linux UI become "better" in the future? Why isn't these things implemented now?
The UI on Linux has been *steadily* (and compared to competitors, extremely rapidly) been improving. About twelve years ago, Linux didn't even exist. About ten years ago, you needed to be a bit of a kernel hacker to consider touching Linux. Seven years ago, a fairly serious techie experimenter, comfortable with poking around with your bootloaders. Five years ago, you had to still be a pretty decent power user, be comfortable not having a GUI for configuring much of anything, and be able to deal with lots of incompatibilities with Windows software, much less little hardware support. Four years ago, you had to be willing to deal with pretty alpha-ish, flaky or archaic desktop environment software, and still had to worry pretty constantly about hardware compatibility.
Frankly, Linux as a general user desktop environment has essentially gone from zero to threat #1 on Microsoft's worry list in the last three or four years. In some areas, UIs on Linux have surpassed their Microsoft equivalents. KDE's use of detachable panes or GNOME's complete user-configurability of keyboard menu equivalents are pretty neat. Four years ago, Linux multimedia was a pretty sad thing -- there was a commercial mpeg player called mtv and a couple of projects. Today, properly set up Linux boxes smoke Windows in latency. Microsoft has not adapted will to the tougher security requirements of an Internet-connected age, as Linux has.
Linux still has issues that keep some people from using it. A lack of entertainment software (most traditional video games do not make very good open source projects) is significant. Poor inter-distro binary and library compatibility is also an issue. If I had to ship something in binary format that I knew would run on Linux boxes, I'd probably ship it in PE format, because Wine can provide stronger guarantees about binary compatibility than Linux itself can. Linux also does not currently, IMHO, cater as well to the power-user-but-not-techie as Linux does. The light user, who uses a spreadsheet, word processor, email program, and web browser (oh, and Solitaire), has little problem with Linux other than an inability to interact with Microsoft Office file formats reliabily, and enjoys increased stability. The techie loves Linux's ability to be remotely administered, its performance, customizability, scriptability, huge (and free) suite of development tools, and availability of source to fix irritating bugs. The almost-techie-power-user, however, runs into problems. Linux has a thinner layer of GUI over the internals than does Windows. They're probably going to have to interact with the CLI. The power user may want to install unusual software, the sort of thing that doesn't come packaged, but be incapable of dealing with any problems in compiling that software.
So I believe that Linux is getting better for most desktop users much faster than Windows is, but there are definitely categories of users that will not be happy with Linux.
Yeah, Ikaruga is neat, but it's really a different sort of game from what I'm talking about. The traditional top-down shooter has huge, crazy, powerupable weapons (Ikaruga you have your special weapon, but that's it). Also, Ikaruga is a bit too hard for me. It's a really tough game.
Thanks for the links -- I knew about zinf, which is the old freeamp renamed (IIRC, development on it sort of foundered) The JuK website is certainly...brightly colored.
Mike, this looks very interesting indeed. If this is what I think it is, and it gets the support it needs, it will finally fix a longstanding and huge hole in Linux desktop usability.
Just so that I'm sure (I've skimmed a few Google searchs, and am still not sure), is this a fallback-capable system, or just a mixer with an ALSA interface? Does it use hardware channels as available until out of them and *then* do software mixing, or is it just another software mixer?
So. What's the damn use of having a PCI-X bus backwards compatible in therms of clock frequency if it's volt incompatible?
Probably due to the fact that it's pretty easy to change voltage, but a pain in the ass to generate a host-synchronized, extremely precise different clock signal. So it's easy to wrap an existing 5V system in a 3.3V card. It's a wee bit more difficult to take existing circuitry and redesign it to work at a different clock rate.
It doesn't affect users directly, just engineering firms.
So why are you using IE?