What exactly have you been doing with your shell? I have never had anything that approaches this sort of problem; on a few occasions I wind up forgetting which of a handful of columns from the output of "ps" or "ls -l" is the one I want to sort by.
Hmm, some examples looking back through my history:
I've got things like:
grep processor/proc/cpuinfo | cut -d: -f2 | tail -n1 to extract the number of processors (after which I decided it would be better to replace the cut/tail with wc -l). You could replace cut -d: -f2 with something like select Number.
There was the time I tried to get a list of all files that were transitively included by some file, so I tried a few variations of gcc -fsyntax-only -H sharedptr.cc | cut -f2 "-d " before giving up and using sed. If GCC gave objects as output, I could have said select FileName or something instead.
There was the time I wanted to send a signal to some occurrences of a process, so I had something like ps | (some greps) | cut -c10-15 | xargs -n1 kill -sCONT.
I sort of feel like "replace cut" is a typical example of this, but it's not the only one. I'm not sure what else to look for in my history though.
you are still stuck having to remember dozens of fields for whatever class you are dealing with.
Which is easier? Remembering "the PID is in the column named
Pid
" or remembering "the PID is in columns 10-15. Oh, unless the width of the first column depends on the maximum username length being displayed, in which case it might change."
And even if you forget, you could run ps and it would tell you, as opposed to you having to count fields.
Do you have an actual example that you could share? Again, this is a problem that I have never encountered, and I have been using GNU for a long time.
If you want a full discussion of the problems you can face, see here. (My executive summary is that "if you've used xargs, you've probably used something that is broken -- or at least not general.")
If VS "sucks" for C++, I'm not sure what works well. Probably nothing. Have any suggestions?
(It certainly doesn't have the level of support that C# has, but IMO VS works way better for C++ than anything else I've tried. In particular, I think it generally works way better than Emacs + Bash + GCC + etc.)
You don't need to learn anything new in order to get on with it
Uh, if you already know it.
while unix uses the same filesystem APIs for both.
Meanwhile chmod modifies traditional Unix permissions, while acl_* functions modify the ACLs. From the user's perspective, chmod and ls -l is used for traditional permissions, but setfacl and getfacl are used for ACLs. I'm not sure why you're trying to argue that Unix is consistent on this matter...
Unix permissions are less flexible, but provide everything that 99% of users require.
Whereas if they don't, you're totally screwed. Want to share a file with just one or two other people, but don't have root (to create a group)? Have fun with traditional Unix permissions.
(In undergrad, when I worked on a group project, our security was "gee, I hope no one guesses the name of this directory.")
The biggest problem is that you often need to parse (and reparse, and reparse...) data because.... it's in a textual format. Most of the time it's easy parsing, like extracting a column, but it's still obnoxious. ("Does it start at column 40? No? 45? No? How' bout 43?" Or you write a "ruler" script.)
Parsing file names in particular is... "interesting". It's basically never worth it to get it actually correct, which should tell you that something in the toolchain is doing it wrong.
And yet, for all of that, the SH-variants have an enormous body of code behind them
And yet, Windows has lots of legacy apps despite it being in many ways inferior to *nixs. (I believe in the last bit of that rather less than your typical/.er.) Popularity isn't particularly good evidence of not being sucky.
it's a fucking nightmare to code in
As opposed to sh?...but rather that you could automate tasks, and at that, the sh family works remarkably well, and has done so for decades
And a more PS-like shell could do it far better in many cases.
(Okay, in some sense I'm lumping way too much in with the shell, and am considering the common utilities like ls in there too. Really, a new suite of utilities without a new shell could do more than a new shell with the same old utilities.)
I would love it if a properly integrated bash variant was available on Windows. Not one that requires some awful layer like Cygwin, but one that runs natively without some ghastly compatibility layer.
Personally, I think PowerShell is a lot closer to the ideal shell for today than Bash is (and I'm typing this on Linux). PS is kind of maddening to use because of some things like the tab completion differences (I've tried to give it a fair shot, but I really don't like it) and the god-awful "terminal emulator" that it runs in.
But I strongly feel that if the Linux folks would take a step back and acknowledge that it's no longer 1970, they'd see that have programs set up to pass objects around instead of text can be hugely beneficial.
(I'm open to some textual serialization of objects, such as JSON or similar.)
NT has been closed-source and multi-arch since it's inception, basically. What architectures it's supported has changed over time, but at one time or another it's run on x86, x64, Power, MIPS, Alpha, Itanium, and now ARM.
Handwriting is still an inferior input mechanism to keyboards...
That's true in most situations, but I strongly feel that it's wrong for note-taking. I've been a grad student for a while -- I've taken a lot of classes. And I've tried the "type notes" thing on a few occasions, and never really liked it. Why? Because while the keyboard is the best device out there at generating text, there's a crapload of stuff that is useful to put into notes that isn't text. A keyboard is awful at diagrams. (Unless you've got an hour to lay it out with PGF/Tikz or something.) An actual mouse would be okay but not great, and the mouse-replacements you get with laptops were bad.
I always went back to pencil and paper because it was the superior input mechanism.
I got a tablet after a year or two in grad school, and it was absolutely excellent for class note taking. In some ways it wasn't quite as nice as paper, but it made up for it in searchability and such. (A little contrary to what Jobs says, MS's handwriting recognition is decent in general and excellent in OneNote. (I might be wrong, but I have some evidence that I think strongly suggests that OneNote's is really good because it doesn't have to commit to any one interpretation. If you search for 'foo' then search for 'bar', and you have a squiggle that looks a bit like each, it can point you at it in both cases. It is much worse if you actually do a handwriting->text conversion.))
If I didn't stop taking classes a year or so after getting the tablet, I have no doubt that I would have continued using it. It's way better.
That being said, it's probably less of a boon in something like a corporate meeting scenario, and "students taking classes" is a relatively nitche and pretty poor overall population -- even if tablets work great in it.
(Yes, I know that some people use laptops to take notes by typing. Most of them don't -- even though they own them. I think that says something. And like I said, I hated it.)
I think you're confusing Win95 and the 3.x (and earlier) series. Unless you're a very very strange person, you didn't get to the Windows 95 desktop through DOS.
It's definitely not for everything, at least not yet. This page has a table with a bunch of different kinds of devices and the appropriate API for them. If I counted right (entirely possible I didn't), there are 9 out of 30 rows where UMDF is at least sometimes available; in 3 cases, it's the only choice.
OK, looks like I may have been wrong about my comparison with Linux; it seems its support for user-mode drivers is better than I thought. Still, it's not new to Windows either.
...even Microsoft starts to understand that modularity is the way to go while designing complex systems. Moving various operating system components to the user space is just a logical conclusion of the research done during the last four decades.
MS has understood that longer than you think; in fact, Windows is rather better in that regard than Linux is. Vista in particular was a big turning point with the introduction of the usermode driver framework (UMDF), which put a lot drivers in userspace. (I'm not sure of the details, e.g. whether the UMDF is the only option if you're writing such a driver.) Heck, the first version of NT back in 1990-whatever even put the graphics driver in usermode: if your graphics driver crashed, the system would just restart it. (Graphics drivers were moved into the kernel for performance reasons and remain there now.)
As for explorer, I don't think it's ever run in kernel mode. It's always been "just another app" from the system's perspective. You can even replace explorer with another desktop environment if you'd like; I remember running Litestep back in Windows 98.
What this article is about is the user's perspective. The standard desktop is no longer going to be the first thing you see when you turn on or log onto your computer, and you'll have to explicitly start it.
(And their new "tile" thing will continue to run in userspace.)
Falling off of a playground gym won't result in a crater the size of a medium-sized city.
And just to be clear, I'm talking about the larger objects the second article mentions. I think trying to capture the smaller objects highlighted by the paper is probably fine.
I give one way of building up to the goal (use Mars and Venus); size is an orthogonal way. I don't think that trying a large object around Earth is a good idea before Mars/Venus, even if you build up with smaller objects first.
Falling off of a playground gym won't result in a crater the size of a medium-sized city.
I'm all for crazy sci-fi things, and I lament the US's turning away from NASA. But at the same time, take things in reason. Bringing asteroids in "close" to earth might be a good goal for a couple decades out, but not now. Have a couple practice runs with Mars and Venus first.
The US didn't say "ok, we're going to the moon now" and have their first launch be to the moon, we built up to it. Hell, the first two launches weren't even orbital.
The article isn't very specific in terms of how the plan maps out though.
if you don't like the way the OS is going then why would you even use it?
Because if I took your suggestion, I wouldn't have a computer?
Less flippantly, there are more things to an OS than the file manager. It's perfectly reasonable that you'd like some of the other changes despite the file manager and see the upgrade as a net positive.
I use Windows at home, mostly by choice. Why? Because I use Linux at work, and I hate both of them, and it's nice to be pissed off for different reasons at work and at home.
There's plenty to like about Windows, and a lot to dislike about Linux.
Eh, only sort of. Everything that explorer needs to read in order to get the estimated time it will need to read when it copies -- and then when it goes to copy, it's already in the OS's buffer cache, so it's just moving the work around.
The exception to this is if the copy blows your buffer cache and flushes the otherwise-preloaded MFT portions. And in that case, the copy is going to take forever anyway, so taking 2*forever isn't a huge deal.
Personally, if someone writes a 'cp' or 'mv' on Linux that will do the same and then give me a progress indicator, I'll switch to it immediately. (Anyone know of one? I've put some time into it, but the closest I've seen is one that just gives you a progress bar for single large files.) IMO the only time that the initial gathering phase takes an appreciable amount of time is when you are going to really want the progress bar anyway.
TLDR: You say "bigger problem", I say "wonderful feature".
I'd have modded you up except I'd rather nitpick.:-)
Oh and here's a little known fact: DOS (and Windows) allows the user to use either \ or / as a path separator.
This is true... sort of. The problem is it's not completely consistent from a UI perspective, even among MS stuff. Lots of CMD's builtin commands won't work with / separators, and it won't tab complete through them. ('dir foo/bar' gives 'paramater format not connect'.)
The standard file dialogs also don't work properly if you give them a path with/s, nor do they offer completions.
Slashdot Mirror
What exactly have you been doing with your shell? I have never had anything that approaches this sort of problem; on a few occasions I wind up forgetting which of a handful of columns from the output of "ps" or "ls -l" is the one I want to sort by.
Hmm, some examples looking back through my history:
I've got things like:
I sort of feel like "replace cut" is a typical example of this, but it's not the only one. I'm not sure what else to look for in my history though.
you are still stuck having to remember dozens of fields for whatever class you are dealing with.
Which is easier? Remembering "the PID is in the column named
" or remembering "the PID is in columns 10-15. Oh, unless the width of the first column depends on the maximum username length being displayed, in which case it might change."
And even if you forget, you could run ps and it would tell you, as opposed to you having to count fields.
Do you have an actual example that you could share? Again, this is a problem that I have never encountered, and I have been using GNU for a long time.
If you want a full discussion of the problems you can face, see here. (My executive summary is that "if you've used xargs, you've probably used something that is broken -- or at least not general.")
If VS "sucks" for C++, I'm not sure what works well. Probably nothing. Have any suggestions?
(It certainly doesn't have the level of support that C# has, but IMO VS works way better for C++ than anything else I've tried. In particular, I think it generally works way better than Emacs + Bash + GCC + etc.)
You don't need to learn anything new in order to get on with it
Uh, if you already know it.
while unix uses the same filesystem APIs for both.
Meanwhile chmod modifies traditional Unix permissions, while acl_* functions modify the ACLs. From the user's perspective, chmod and ls -l is used for traditional permissions, but setfacl and getfacl are used for ACLs. I'm not sure why you're trying to argue that Unix is consistent on this matter...
Unix permissions are less flexible, but provide everything that 99% of users require.
Whereas if they don't, you're totally screwed. Want to share a file with just one or two other people, but don't have root (to create a group)? Have fun with traditional Unix permissions.
(In undergrad, when I worked on a group project, our security was "gee, I hope no one guesses the name of this directory.")
The biggest problem is that you often need to parse (and reparse, and reparse...) data because.... it's in a textual format. Most of the time it's easy parsing, like extracting a column, but it's still obnoxious. ("Does it start at column 40? No? 45? No? How' bout 43?" Or you write a "ruler" script.)
Parsing file names in particular is... "interesting". It's basically never worth it to get it actually correct, which should tell you that something in the toolchain is doing it wrong.
And yet, for all of that, the SH-variants have an enormous body of code behind them
And yet, Windows has lots of legacy apps despite it being in many ways inferior to *nixs. (I believe in the last bit of that rather less than your typical /.er.) Popularity isn't particularly good evidence of not being sucky.
it's a fucking nightmare to code in
As opposed to sh? ...but rather that you could automate tasks, and at that, the sh family works remarkably well, and has done so for decades
And a more PS-like shell could do it far better in many cases.
(Okay, in some sense I'm lumping way too much in with the shell, and am considering the common utilities like ls in there too. Really, a new suite of utilities without a new shell could do more than a new shell with the same old utilities.)
I would love it if a properly integrated bash variant was available on Windows. Not one that requires some awful layer like Cygwin, but one that runs natively without some ghastly compatibility layer.
Now I actually agree with you there.
Personally, I think PowerShell is a lot closer to the ideal shell for today than Bash is (and I'm typing this on Linux). PS is kind of maddening to use because of some things like the tab completion differences (I've tried to give it a fair shot, but I really don't like it) and the god-awful "terminal emulator" that it runs in.
But I strongly feel that if the Linux folks would take a step back and acknowledge that it's no longer 1970, they'd see that have programs set up to pass objects around instead of text can be hugely beneficial.
(I'm open to some textual serialization of objects, such as JSON or similar.)
"going to be"?
NT has been closed-source and multi-arch since it's inception, basically. What architectures it's supported has changed over time, but at one time or another it's run on x86, x64, Power, MIPS, Alpha, Itanium, and now ARM.
Handwriting is still an inferior input mechanism to keyboards...
That's true in most situations, but I strongly feel that it's wrong for note-taking. I've been a grad student for a while -- I've taken a lot of classes. And I've tried the "type notes" thing on a few occasions, and never really liked it. Why? Because while the keyboard is the best device out there at generating text, there's a crapload of stuff that is useful to put into notes that isn't text. A keyboard is awful at diagrams. (Unless you've got an hour to lay it out with PGF/Tikz or something.) An actual mouse would be okay but not great, and the mouse-replacements you get with laptops were bad.
I always went back to pencil and paper because it was the superior input mechanism.
I got a tablet after a year or two in grad school, and it was absolutely excellent for class note taking. In some ways it wasn't quite as nice as paper, but it made up for it in searchability and such. (A little contrary to what Jobs says, MS's handwriting recognition is decent in general and excellent in OneNote. (I might be wrong, but I have some evidence that I think strongly suggests that OneNote's is really good because it doesn't have to commit to any one interpretation. If you search for 'foo' then search for 'bar', and you have a squiggle that looks a bit like each, it can point you at it in both cases. It is much worse if you actually do a handwriting->text conversion.))
If I didn't stop taking classes a year or so after getting the tablet, I have no doubt that I would have continued using it. It's way better.
That being said, it's probably less of a boon in something like a corporate meeting scenario, and "students taking classes" is a relatively nitche and pretty poor overall population -- even if tablets work great in it.
(Yes, I know that some people use laptops to take notes by typing. Most of them don't -- even though they own them. I think that says something. And like I said, I hated it.)
I think you're confusing Win95 and the 3.x (and earlier) series. Unless you're a very very strange person, you didn't get to the Windows 95 desktop through DOS.
It's definitely not for everything, at least not yet. This page has a table with a bunch of different kinds of devices and the appropriate API for them. If I counted right (entirely possible I didn't), there are 9 out of 30 rows where UMDF is at least sometimes available; in 3 cases, it's the only choice.
Maybe I'm wrong, but when explorer.exe is closed, you don't see ANY windows until it is restarted.
You're wrong.
All your windows stay open (and accessible through, e.g., alt-tab).
OK, looks like I may have been wrong about my comparison with Linux; it seems its support for user-mode drivers is better than I thought. Still, it's not new to Windows either.
...even Microsoft starts to understand that modularity is the way to go while designing complex systems. Moving various operating system components to the user space is just a logical conclusion of the research done during the last four decades.
MS has understood that longer than you think; in fact, Windows is rather better in that regard than Linux is. Vista in particular was a big turning point with the introduction of the usermode driver framework (UMDF), which put a lot drivers in userspace. (I'm not sure of the details, e.g. whether the UMDF is the only option if you're writing such a driver.) Heck, the first version of NT back in 1990-whatever even put the graphics driver in usermode: if your graphics driver crashed, the system would just restart it. (Graphics drivers were moved into the kernel for performance reasons and remain there now.)
As for explorer, I don't think it's ever run in kernel mode. It's always been "just another app" from the system's perspective. You can even replace explorer with another desktop environment if you'd like; I remember running Litestep back in Windows 98.
What this article is about is the user's perspective. The standard desktop is no longer going to be the first thing you see when you turn on or log onto your computer, and you'll have to explicitly start it.
(And their new "tile" thing will continue to run in userspace.)
Falling off of a playground gym won't result in a crater the size of a medium-sized city.
And just to be clear, I'm talking about the larger objects the second article mentions. I think trying to capture the smaller objects highlighted by the paper is probably fine.
I give one way of building up to the goal (use Mars and Venus); size is an orthogonal way. I don't think that trying a large object around Earth is a good idea before Mars/Venus, even if you build up with smaller objects first.
Falling off of a playground gym won't result in a crater the size of a medium-sized city.
I'm all for crazy sci-fi things, and I lament the US's turning away from NASA. But at the same time, take things in reason. Bringing asteroids in "close" to earth might be a good goal for a couple decades out, but not now. Have a couple practice runs with Mars and Venus first.
The US didn't say "ok, we're going to the moon now" and have their first launch be to the moon, we built up to it. Hell, the first two launches weren't even orbital.
The article isn't very specific in terms of how the plan maps out though.
Yeah, agreed. The summary says "at first glance" it looks like a bad idea, but to me it looks like a bad idea the next few glances as well.
Try it out with Mars and Venus a couple times first... then we can talk.
I seem to have accidentally a word.
"On what planet does Windows center any application title?"
THEN we get the application title - but oh no, it's not centred.
On what planet does Windows any application title?
...and if you "streamline" your W8 explorer, you'll get those 6 back too. Whoopdie do.
if you don't like the way the OS is going then why would you even use it?
Because if I took your suggestion, I wouldn't have a computer?
Less flippantly, there are more things to an OS than the file manager. It's perfectly reasonable that you'd like some of the other changes despite the file manager and see the upgrade as a net positive.
I bet you hate menus, then.
A hate to disappoint, but the moon is clearly out.
I use Windows at home, mostly by choice. Why? Because I use Linux at work, and I hate both of them, and it's nice to be pissed off for different reasons at work and at home.
There's plenty to like about Windows, and a lot to dislike about Linux.
Eh, only sort of. Everything that explorer needs to read in order to get the estimated time it will need to read when it copies -- and then when it goes to copy, it's already in the OS's buffer cache, so it's just moving the work around.
The exception to this is if the copy blows your buffer cache and flushes the otherwise-preloaded MFT portions. And in that case, the copy is going to take forever anyway, so taking 2*forever isn't a huge deal.
Personally, if someone writes a 'cp' or 'mv' on Linux that will do the same and then give me a progress indicator, I'll switch to it immediately. (Anyone know of one? I've put some time into it, but the closest I've seen is one that just gives you a progress bar for single large files.) IMO the only time that the initial gathering phase takes an appreciable amount of time is when you are going to really want the progress bar anyway.
TLDR: You say "bigger problem", I say "wonderful feature".
I'd have modded you up except I'd rather nitpick. :-)
Oh and here's a little known fact: DOS (and Windows) allows the user to use either \ or / as a path separator.
This is true... sort of. The problem is it's not completely consistent from a UI perspective, even among MS stuff. Lots of CMD's builtin commands won't work with / separators, and it won't tab complete through them. ('dir foo/bar' gives 'paramater format not connect'.)
The standard file dialogs also don't work properly if you give them a path with /s, nor do they offer completions.