Well, if you want really bug-free software, take a look at how the Space Shuttle software engineers do it..
That is an excellent article, one of the best that I have ever read. The process, not the code is the software.
Unfortunately, this will likely not work for general purpose software companies. How do you define the requirements of software when you don't know exactly who your customers are, and what they want the software to do. Or, what other software will be interacting with your software, and whether the underlying OS is operating correctly?
The space shuttle software engineers have the luxury of living in a virutal bubble, where they know every single detail of the platform they are running on, and have users who are technically adept and know exactly what they want . Most general purpose developers do not have that luxury, so there will always be errors. Don't get me wrong, every developer can learn a lot from the way these guys do business, but their solution is not a panecea to the software bug problem.
The description above makes this book sound a lot like the Robert X. Cringely show on PBS a few years back (the name escapes me now).
I wonder if this book is just a rehash of that show or if there is a lot of new information.
If my understanding is correct, it is not SMP you are talking about, but AMP or AsMP, as in Asymmetric
Not correct. Ever since OS 8.6 the MacOS Nanokernel has been SMP. The macos runs as one thread on the nanonkernel (called the BlueBox) and is scheduled along with other MP threads.
This is why it was so easy to make Classic work on X. Except on X the Blue Box (Classic) is one pthread ( a direct mapping to a Mach kernel thread) on top of Mach/BSD.
Full booting without the GUI is not possible out of the box. Either you boot single user, and about nothing is configured (as expected), or you boot verbose, but end up in the GUI (from which you can go text mode by using '>console' as a user name).
To get rid of all the gui, you had to change/etc/ttys on NeXTstep / Mac OS X Server. Probably the same under OS X. In that case, all the services are started, as expected.
That still looks to be the case. From OSX PB/etc/ttys:
# @(#)ttys 5.2 (Berkeley) 6/10/93
#
# name getty type status comments
#
# If the console is marked insecure, single-user requires
# the root password.
#console "/usr/libexec/getty std.9600" vt100 on secure
console "/System/Library/CoreServices/loginwindow.app/logi nwindow" vt100 on secure window=/System/Li
brary/CoreServices/WindowServer onoption="/usr/libexec/getty std.9600"
So it looks like you would just un-comment the first console, and comment the second one to boot into text mode.
The other question I have regarding UFS vs. HFS+ on OS X, is, if you upgrade a currently formatted HFS+ partition, can you change it to UFS, or are you stuck with HFS+ (unless you reformat)?
You are "stuck", as you put it, with HFS+ unless you reformat. Of course if you do use UFS (FFS), then you cannot use Classic, because Classic requires HFS+.
And if you DO stay with HFS+, can it support traditional Unix file permissions, or are you screwed, and if you DON'T have Unix file permissions, how do you secure your fs from intruders?
No, you are not screwed, HFS+ has full support for UNIX style permissions. Mac OS Classic just ignores any set permissions.
BTW, another benifit of using HFS+, is that you will have real Aliases, instead of SymLinks. Aliases will not break when you move the original, SymLinks do.
There is no real benifit to using UFS over HFS+ unless you absolutly need case sensitive file names (eg. to build Perl).
It's in the process of being synced with the most current BSD, though it is written as a user-mode process which runs on top of Mach. You could run several such processes simultaneously. In short, no, the NeXT stuff wasn't dumped at all...it's just totally revamped.
Couple of corrections: 1. s/BSD/FreeBSD, 2. Mach and the BSD kernel live in the same address space. The BSD kernel still calls all of the Mach interfaces, but does so directly and not through Mach messages as a user process. The "BSD as a user process" was brought up in the BSD kernel session at WWDC this year, and the Apple guys made sure everyone understood that BSD and Mach share the same kernel address space for performance reasons.
Not quite. Mach and the BSD kernel on OSX live in the same kernel address space. While the BSD kernel does use the Mach interfaces, it is a direct call, and not done through Mach IPC. If there is a speed hit at all in MacOSX it will be very minor.
This explanation comes straight from Apple, I was in attendance at the WWDC Mac OS X kernel session. They (Apple) realized that using a pure MicroKernel was great for academic research, it was not great for a Consumer OS.
Some of us write code that runs under Linux, but does not run under MacOS X.
What the hell are you talking about?! You would have to write some seriously messed up code (or be writing a kernel module) to not be compatible with all UN*X like OS's. I can take just about any GNU/Linux/UN*X program out there, make a few trivial changes (mostly to autoconf) and recompile on Mac OS X Server just fine. And with Mac OS X having feature parity with FreeBSD, this will be even easier.
I want to know where the obligatory poll is that usually accompinies articles covering "hot topics". You know, something like the following; short and sweet.
" Splitting MS up will cause more viruses. Do you aggree with Bill Gates?"
Aggree, ILOVEYOU would have been worse without MS's great software.
Disagree, this is just more MS fud from the master.
Most likely ol' Bill vetoed the poll, because even he knows that 99.8% of voters would have disagreed. I just cannot believe the audacity of Bill Gates, does he think that all computer users are MS lemmings?
I am wondering if this book is really necessary. Since Alan Cox and other core kernel developers are working on commenting the kernel themselves I don't really see the usefulness of this book.
Personally I would much rather learn how the kernel works from the actual kernel developers than anyone else. ( I don't know if Mr. Maxwell contributes to the kernel, but I have never heard of him.)
If you want a paper version (which is sometimes better then a computer screen) just write a little script that recursively descends the kernel directory hierarchy and prints each *.h and *.c file.
I think Apple is already thinking about how to get QT on other platforms (besides Mac OS and Windows). I think their answer will be QuickTime for Java. With QT 4.1, Java applets can now make calls to the QuickTime sub-system. I don't know how much of the API is available through Java, but this will most likely be how Apple achieves true cross-platform support. Check out the QuickTime for Java book for more info..
As to Apple having a QT player with an X (as in X Window, not Mac OS X) interface floating around, that is very unlikely. Apple's GUI for OS X will be there own vector based Quartz. Apple will not officially be supporting X Windows on Mac OS X (although John Carmak has ported XFree86 to Mac OS X).
One last thing, the Sorenson codec is not Apple's to Open Source, it is licensed from a third-party, as are many parts of QuickTime. This of course, would be a major hinderance to Open Sourcing QuickTime.
Imagine, for example, a browser that, while a given page is loading, goes out and pings the pages that are linked to, checks their existence and the time it takes to get to them, and displays this information when you roll over the link.
Better yet, a browser that retrieved all links on a page that matched a set of user defined keywords. That way when the user clicked on a link the page would be pulled from the browser's cache. You could even have a setting that determined how deep the browser would follow links.
Granted this would generate more traffic, but most people sit idle while reading a web page and this would help a lot in the perceived slowness of the Internet.
Well, it can't. There is the option of enabling SMP support when you do a make xconfig, but enabling it doesn't actually do anything.
Compile times for a Dual PII 450 w/448MB RAM. Resulting kernel was 650K. 'jx' is the number of make's running, s means single CPU. egcs-2.91.66:2.2.13:j4 261.180u 20.280s 2:27.88 190.3% 0+0k 0+0io 334324pf+0w gcc-2.7.2.3:2.2.13:j3 232.540u 20.200s 2:11.95 191.5% 0+0k 0+0io 265191pf+0w gcc-2.7.2.3:2.2.13:s 261.76user 16.23system 4:38.48elapsed 99%CPU
What's really interesting is how people claim that their multi-CPU machine is "so much faster" when they enable SMP. Isn't that what psychiatrists call a "self-serving fantasy?"
Damm! That is some fantasy I have been having. I must have been dreaming when my kernel compile times were cut by more than 50%.
You should only have to pay $5K if you take all of RedHat's training classes (4 total). And then the price would be pretty much inline with MSCE and CNE training costs.
The actual RedHat certification test (RH302) is $750. If your people are already knowledgeable about UNIX, but not necessarily Linux or RedHat, then the best option is to send them to RH300 which is a week long "rapid track" training class that includes the test. RH300 is $2700.
As far as locations, now that RedHat has partnered with GlobalKnowledge, you should be able to find a training center fairly close to you. See http://am.globalknowledge.com/redhat/index.html for more info.. A quick check at the GK site shows that RH302 is being offered in TX, DC, MN, CA, MA, NJ and CO in the coming months.
BTW, I don't work for RedHat or GlobalKnowledge. It just so happens that I recently went through (and passed) RH300.
Looks like Harry Browne's page uses xhtml transitional - with all the lowercase tags.
Sorry, using lowercase tags, does not mean you are using XHTML. Browne's site does not even have a DTD specified, there is no way it can be XHTML compliant.
Incidentally, anyone know where a good comprehensive xhtml tutorial would be? Or a good xhtml WYSIWYG editor?
Well you might want to check out this site to see some actual XHTML. It was mentioned in the "Want to see some good HTML?" post that was moderated off-topic. Don't know why the post was moderated as off topic though, I thought good HTML design was the topic. Also look at the W3's site for more info..
If you want to execute 68k code fast, the most sensible option is to make a dynamically recompiling emulator running on Alpha (or x86)
I'm not sure how true that is. I'm pretty sure PPC chips have some special instructions and such that make it possible to emulate 68k code fairly fast. It's not full hardware emulation or anything, but even old, old powermacs run 68k code at fairly suprising speeds.
Actually the PPC has no instructions to help the Mac execute 68k instructions. The PPC instruction set is a super/sub-set (depending on how you look at it) of IBM's POWER instruction set, which was invented in the late 80's early 90's, long before Apple was looking at the PPC as a replacement for the 68k series. Later, when the G3 was being developed, IBM/Motorola did enhance the chip to take advantage of the way the MacOS works, but the changes were to the pipelining and instruction scheduling, not the instruction set.
The Mac's 68k emulator, implements a 68040's instruction set, minus the floating point/MMU instructions. This is why the emulator identifies itself as a 68020, because the emulator is not a true 68040.
The first generation PPC Macs, the x100's, used an instruction emulator that translated 68k instructions to PPC instructions everytime a 68k instruction was executed. These machines were slower at executing 68k code than the fastest 68k machines (840 and friends).
The second generation PPC Macs introduced the PCI bus, the PPC 604 and a new dynamic emulator that would cache 68k instruction translations. Once an instruction was translated, it never needed to be done again. The PPC 604/132 (forget the model number) was the first PPC that was actually faster at executing 68k code than a 68k machine.
Current G3/4 Macs can execute 68k code at double the speed (or more) of any 68k chip, including Motorola's replacement for the 68040, the 68060. Of course 68k speed with current Macs is really a moot point. All new apps are PPC native, and MacOS 9 is around 70% native, with MacOS X being 100% native and only allowing 68k applications to run inside the backwords compatibility "BlueBox" environment.
>They also requires Javascript, which despite the >fact that it is available only in Netscape or IE, >make also webas dangerous as walking on mines (see >recente CERT advisory)
I know of at least three other browsers that support JavaScript. OmniWeb for Mac OS X, iCab pre1.9 (Mac Only), and Opera which is cross-platform. There are probably others too.
To me, Apple's complaining about MS stealing their look and feel never made sense.
Just a clarification. It was not so much the look and feel of the OS, as it was the technologies behind that look and feel. If you look at the Macintosh Toolbox API and the Windows API, there are a lot of similarities, including the idea of Regions (invented by Jeff Raskin for the original Mac OS) that allow windows to be updated incrementally.
If you saw Pirates of Silicon Valley, the general story was pretty much represented the way it actually happened, although Robert X. Cringely's Triumph of the Nerds has a much better account of what happened. Also check out the Nerds Q&A section for some more info., including a commnet by Jeff Raskin himself.
The MS Natural Keyboard is actually one of the worst ergonomic keyboards. It is arguably better than standard keyboards, but there are better alternatives. One of the best, is the SmartBoard by DataDesk Technologies. It takes a little getting used to, but is well worth the effort. It is also well worth the $100 price tag.
Slashdot Mirror
Omniweb (http://www.omnigroup.com/products/omniweb/) for Mac OS X has and has had all of these features for a while now (except for maybe #3).
That is an excellent article, one of the best that I have ever read. The process, not the code is the software.
Unfortunately, this will likely not work for general purpose software companies. How do you define the requirements of software when you don't know exactly who your customers are, and what they want the software to do. Or, what other software will be interacting with your software, and whether the underlying OS is operating correctly?
The space shuttle software engineers have the luxury of living in a virutal bubble, where they know every single detail of the platform they are running on, and have users who are technically adept and know exactly what they want . Most general purpose developers do not have that luxury, so there will always be errors. Don't get me wrong, every developer can learn a lot from the way these guys do business, but their solution is not a panecea to the software bug problem.
The description above makes this book sound a lot like the Robert X. Cringely show on PBS a few years back (the name escapes me now). I wonder if this book is just a rehash of that show or if there is a lot of new information.
Not correct. Ever since OS 8.6 the MacOS Nanokernel has been SMP. The macos runs as one thread on the nanonkernel (called the BlueBox) and is scheduled along with other MP threads.
This is why it was so easy to make Classic work on X. Except on X the Blue Box (Classic) is one pthread ( a direct mapping to a Mach kernel thread) on top of Mach/BSD.
I'm pretty sure BSA is an acronym for Buttheaded Software Administrators.
To get rid of all the gui, you had to change /etc/ttys on NeXTstep / Mac OS X Server. Probably the same under OS X. In that case, all the services are started, as expected.
That still looks to be the case. From OSX PB /etc/ttys:
So it looks like you would just un-comment the first console, and comment the second one to boot into text mode.You are "stuck", as you put it, with HFS+ unless you reformat. Of course if you do use UFS (FFS), then you cannot use Classic, because Classic requires HFS+.
And if you DO stay with HFS+, can it support traditional Unix file permissions, or are you screwed, and if you DON'T have Unix file permissions, how do you secure your fs from intruders?
No, you are not screwed, HFS+ has full support for UNIX style permissions. Mac OS Classic just ignores any set permissions.
BTW, another benifit of using HFS+, is that you will have real Aliases, instead of SymLinks. Aliases will not break when you move the original, SymLinks do.
There is no real benifit to using UFS over HFS+ unless you absolutly need case sensitive file names (eg. to build Perl).
I don't know if you can buy just Mac OS X server, but you can get it if you buy a server. See http://store.apple.com/1-800-MY-APPLE/WebObjects/A ppl eStore.woa/93/wo/R7NRd027FLzFCXYeu3/2.3.0.3.28.8.1 .23.3.1.1.0?44,5. Or if that link is broken by now (I don't know if WO generates permanant links or not). Go to the Apple Store and then browse to the G4 section, and in the upper right corner there should be a graphic mentioning G4 Servers, click on that, and you should see the X server option.
It's in the process of being synced with the most current BSD, though it is written as a user-mode process which runs on top of Mach. You could run several such processes simultaneously. In short, no, the NeXT stuff wasn't dumped at all...it's just totally revamped.
Couple of corrections:1. s/BSD/FreeBSD,
2. Mach and the BSD kernel live in the same address space. The BSD kernel still calls all of the Mach interfaces, but does so directly and not through Mach messages as a user process. The "BSD as a user process" was brought up in the BSD kernel session at WWDC this year, and the Apple guys made sure everyone understood that BSD and Mach share the same kernel address space for performance reasons.
Not quite. Mach and the BSD kernel on OSX live in the same kernel address space. While the BSD kernel does use the Mach interfaces, it is a direct call, and not done through Mach IPC. If there is a speed hit at all in MacOSX it will be very minor.
This explanation comes straight from Apple, I was in attendance at the WWDC Mac OS X kernel session. They (Apple) realized that using a pure MicroKernel was great for academic research, it was not great for a Consumer OS.
Some of us write code that runs under Linux, but does not run under MacOS X.
What the hell are you talking about?! You would have to write some seriously messed up code (or be writing a kernel module) to not be compatible with all UN*X like OS's. I can take just about any GNU/Linux/UN*X program out there, make a few trivial changes (mostly to autoconf) and recompile on Mac OS X Server just fine. And with Mac OS X having feature parity with FreeBSD, this will be even easier.
I want to know where the obligatory poll is that usually accompinies articles covering "hot topics". You know, something like the following; short and sweet.
" Splitting MS up will cause more viruses. Do you aggree with Bill Gates?"
- Aggree, ILOVEYOU would have been worse without MS's great software.
- Disagree, this is just more MS fud from the master.
Most likely ol' Bill vetoed the poll, because even he knows that 99.8% of voters would have disagreed. I just cannot believe the audacity of Bill Gates, does he think that all computer users are MS lemmings?I am wondering if this book is really necessary. Since Alan Cox and other core kernel developers are working on commenting the kernel themselves I don't really see the usefulness of this book.
Personally I would much rather learn how the kernel works from the actual kernel developers than anyone else. ( I don't know if Mr. Maxwell contributes to the kernel, but I have never heard of him.)
If you want a paper version (which is sometimes better then a computer screen) just write a little script that recursively descends the kernel directory hierarchy and prints each *.h and *.c file.
I think Apple is already thinking about how to get QT on other platforms (besides Mac OS and Windows). I think their answer will be QuickTime for Java. With QT 4.1, Java applets can now make calls to the QuickTime sub-system. I don't know how much of the API is available through Java, but this will most likely be how Apple achieves true cross-platform support. Check out the QuickTime for Java book for more info..
As to Apple having a QT player with an X (as in X Window, not Mac OS X) interface floating around, that is very unlikely. Apple's GUI for OS X will be there own vector based Quartz. Apple will not officially be supporting X Windows on Mac OS X (although John Carmak has ported XFree86 to Mac OS X).
One last thing, the Sorenson codec is not Apple's to Open Source, it is licensed from a third-party, as are many parts of QuickTime. This of course, would be a major hinderance to Open Sourcing QuickTime.
Imagine, for example, a browser that, while a given page is loading, goes out and pings the pages that are linked to, checks their existence and the time it takes to get to them, and displays this information when you roll over the link.
Better yet, a browser that retrieved all links on a page that matched a set of user defined keywords. That way when the user clicked on a link the page would be pulled from the browser's cache. You could even have a setting that determined how deep the browser would follow links.
Granted this would generate more traffic, but most people sit idle while reading a web page and this would help a lot in the perceived slowness of the Internet.
Well, it can't. There is the option of enabling SMP support when you do a make xconfig, but enabling it doesn't actually do anything.
Compile times for a Dual PII 450 w/448MB RAM. Resulting kernel was 650K. 'jx' is the number of make's running, s means single CPU.
egcs-2.91.66:2.2.13:j4 261.180u 20.280s 2:27.88 190.3% 0+0k 0+0io 334324pf+0w
gcc-2.7.2.3:2.2.13:j3 232.540u 20.200s 2:11.95 191.5% 0+0k 0+0io 265191pf+0w
gcc-2.7.2.3:2.2.13:s 261.76user 16.23system 4:38.48elapsed 99%CPU
What's really interesting is how people claim that their multi-CPU machine is "so much faster" when they enable SMP. Isn't that what psychiatrists call a "self-serving fantasy?"
Damm! That is some fantasy I have been having. I must have been dreaming when my kernel compile times were cut by more than 50%.
Think before you speak.You should only have to pay $5K if you take all of RedHat's training classes (4 total). And then the price would be pretty much inline with MSCE and CNE training costs.
The actual RedHat certification test (RH302) is $750. If your people are already knowledgeable about UNIX, but not necessarily Linux or RedHat, then the best option is to send them to RH300 which is a week long "rapid track" training class that includes the test. RH300 is $2700.
As far as locations, now that RedHat has partnered with GlobalKnowledge, you should be able to find a training center fairly close to you. See http://am.globalknowledge.com/redhat/index.html for more info.. A quick check at the GK site shows that RH302 is being offered in TX, DC, MN, CA, MA, NJ and CO in the coming months.
BTW, I don't work for RedHat or GlobalKnowledge. It just so happens that I recently went through (and passed) RH300.
Looks like Harry Browne's page uses xhtml transitional - with all the lowercase tags.
Sorry, using lowercase tags, does not mean you are using XHTML. Browne's site does not even have a DTD specified, there is no way it can be XHTML compliant.
Incidentally, anyone know where a good comprehensive xhtml tutorial would be? Or a good xhtml WYSIWYG editor?
Well you might want to check out this site to see some actual XHTML. It was mentioned in the "Want to see some good HTML?" post that was moderated off-topic. Don't know why the post was moderated as off topic though, I thought good HTML design was the topic. Also look at the W3's site for more info..
I'm not sure how true that is. I'm pretty sure PPC chips have some special instructions and such that make it possible to emulate 68k code fairly fast. It's not full hardware emulation or anything, but even old, old powermacs run 68k code at fairly suprising speeds.
Actually the PPC has no instructions to help the Mac execute 68k instructions. The PPC instruction set is a super/sub-set (depending on how you look at it) of IBM's POWER instruction set, which was invented in the late 80's early 90's, long before Apple was looking at the PPC as a replacement for the 68k series. Later, when the G3 was being developed, IBM/Motorola did enhance the chip to take advantage of the way the MacOS works, but the changes were to the pipelining and instruction scheduling, not the instruction set.
The Mac's 68k emulator, implements a 68040's instruction set, minus the floating point/MMU instructions. This is why the emulator identifies itself as a 68020, because the emulator is not a true 68040.
The first generation PPC Macs, the x100's, used an instruction emulator that translated 68k instructions to PPC instructions everytime a 68k instruction was executed. These machines were slower at executing 68k code than the fastest 68k machines (840 and friends).
The second generation PPC Macs introduced the PCI bus, the PPC 604 and a new dynamic emulator that would cache 68k instruction translations. Once an instruction was translated, it never needed to be done again. The PPC 604/132 (forget the model number) was the first PPC that was actually faster at executing 68k code than a 68k machine.
Current G3/4 Macs can execute 68k code at double the speed (or more) of any 68k chip, including Motorola's replacement for the 68040, the 68060. Of course 68k speed with current Macs is really a moot point. All new apps are PPC native, and MacOS 9 is around 70% native, with MacOS X being 100% native and only allowing 68k applications to run inside the backwords compatibility "BlueBox" environment.
I know of at least three other browsers that support JavaScript. OmniWeb for Mac OS X, iCab pre1.9 (Mac Only), and Opera which is cross-platform. There are probably others too.
Just a clarification. It was not so much the look and feel of the OS, as it was the technologies behind that look and feel. If you look at the Macintosh Toolbox API and the Windows API, there are a lot of similarities, including the idea of Regions (invented by Jeff Raskin for the original Mac OS) that allow windows to be updated incrementally.
If you saw Pirates of Silicon Valley, the general story was pretty much represented the way it actually happened, although Robert X. Cringely's Triumph of the Nerds has a much better account of what happened. Also check out the Nerds Q&A section for some more info., including a commnet by Jeff Raskin himself.
http://www.datadesktech.com