Didn't companies try that towards the end of the dot com boom? I seem to recall that many of those internet applications were supposed to be moving IT responsibilities from the department to that company. How many of them survived?
The problem is that most reasonably sized departments need an IT staff anyway. Having them run a mail server or the like isn't that big a deal. While some things can be effecient to subcontract out (i.e. your web server) often it is easier to have it on sight.
There are exceptions. But I think that only a few IT functions really can reasonably be marketed out. I think IBM's marketing strategy will work - but only for a small niche.
I agree. The biggest problem is Linux clustering. Given IBM's current high end prices unless you need consistent uptime reliability, I don't see what benefit IBM could give. Perhaps if they had some nice libraries, say a port of the LANL math libraries. Then a simple way to send the code to IBM and you specify how fast it will run.
However for any application that is being accessed heavily and consistently it would seem that a Linix cluster or Sun cluster would be much more cost effective.
While it is hardly a revolution certain applications might benefit from this approach. Basically you get processing power on demand. Clearly this isn't going to be used by the secretary, accountant or those who use their computers only for spreadsheets and word processessing.
Those who, for example, might need a rendering farm but only for a short time might benefit. Consider that you only pay for the processing you need. If IBM comes up wiht some clustering software that is good and distributed this might work. However it is clearly aimed at the markets that are already buying very large IBM computers. It won't help, for instance, for the typical internet sever.
Having said that though, what kinds of people are that? The main rendering farms are being used fairly consistently. So for them having a bunch of Suns or equivalent systems is more efficient. They can then just add computers. So who is it that would need this sort of thing?
And if they did try and foist it on the general public it would obviously fail immediately. After all the heaviest users of processing time on general computers are games. And most people aren't quite willing to pay for the processing the latest Halo or equivalent might use. (Not to mention the fact that Dell, Apple, and Compact wouldn't follow IBM)
I think it was more just to compare Linux and OSX on equivalent hardware. i.e. how does the Unix underpinnings of Apple compare.
There are problems with this, not the least of which are the silly benchmarks the author used and the apparent lack of optimizations. Further anyone running Linux will either be running it on faster x86 hardware or is running it on a Mac solely because they have old hardware not fast enough to run OSX.
So you are right. Even beyond the benchmark problems it ends up being of limited relevance for deciding on a server.
You do know Alpha is dead. Right? Further the main market Apple is focusing in on are servers for people with OSX in their enterprise. For those types of users XServe is very nice. The fact that it is already competitive with equivalent offerings from DELL is nice.
Yeah most people here would rather roll their own hardware/software system with either Linux or BSD on an x86 system. For larger systems Sun is the reasonable choice. But just because it doesn't meet every need isn't to say that it can't be successful for a server. After al there are different kinds of servers. You wouldn't compare a powerful multiprocessor Sun against Linux on a Pentium4, for instance. They simply are focused in on different markets.
Getting back to your hardware point. Other than focusing in on obsolete hardware, you do have a point. Right now the G4 and chipset Apple uses isn't competitive. Its amazing they got the performance/dollar they did. However with the 970 next year Apple should be able to move into being far more competitive.
Yeah there is always the danger of "vaporware" comparisons. (Remember when the G5 was just around the corner?) But I think IBM is far more reliable than Motorola. Further IBM likely wants to use the chip in their own Linux multiprocessor systems. In that case we'll have a very interesting comparison between OSX and Linux. Further I suspect that while the 970 will initially be middle of the pack, that it may end up doing quite well in multiprocessor servers. By that time OSX and their server software will be in their next major revision and should be far more mature. Hopefully the new filesystem will even be out.
Actually XServe does have a remote GUI admin tool. It isn't quite the same as X11 in terms of generality. I'm told it is very fast though and makes it VERY easy to manage several XServe systems at once. (i.e. a full rack worth of rendering engines)
Of course if you really need full remote control there are all sorts of programs like that for both XP and OSX. Of course they are much slower than X11 whether running on OSX or XP) But then you typically don't need that functionality much. Further those using them probably are running across a 1GB Ethernet communication rather than across the Internet.
That's true in general, but the kinds of services XServe is designed for isn't those markets. In those cases you really should be running a Sun or some custom Linux/BSD servers with the latest chips and a hell of a lot of memory. I believe that XServe is more designed for small businesses or for racks for internal use. (i.e. rendering farms, file servers, intranets, etc.)
In those cases the ease of management will likely be a bigger factor than overall speed. If overall speed is that important you really need someone who is an expert in Linux or BSD, knows how to eck out every bit of speed, knows security up the yazoo, etc. And of course your department will have to pay for this guy. And that ends up being near $100,000 a year once you start including benefits etc. (Or more if you have several folks)
For a more regular IT staff who doesn't want to spend that much time or effort XServe, like the equivalent offerings from DELL, Sun or Compact, is a nice server. As I said for ASP I still think Sun might be a better choice. But it really depends upon what you are doing.
One thing to remember is that OSX is still using the HFS+ file system. The Byte article didn't mention is they were comparing it with the UFS. In either case Apple has new file systems coming down that might shake things up a bit in terms of both reliability and speed. Of course Linux has those now. But I think everyone agrees that XServe is nice for a first generation product. However as with most first generation products, think twice about being on the cutting edge. A better comparison might be a year from now when OSX is up to 10.3 with the equivalent XServe. Hopefully by then the new hardware based on the 970 will be out. Of course by the same measure the Linux/x86 world will also have moved on.
It depends upon what you are comparing it to. According to most benchmarks I've seen when compared to the equivalent offerings from Dell it ends up being cheaper and faster.
I admit I'm a tad skeptical of the relevance of this benchmark, but it does seem that Apple has a nice system. I suspect you could roll your own better with OpenBSD or Linux and a nice AMD multiprocessor system. That's just me though. And realistically a lot of businesses DON'T want such systems. They want a "come as it is" system. Further a lot of people don't want all the messing around that you have to do with most Linux of BSD distributions. Apple has put a very nice interface on their server. Yet you have the added benefit of being able to drop to Unix when necessary.
Apple's big problem is still the chipset used with the G4's. Given that, despite many of the nice features, unless you are primarily serving other Macs, I don't think XServe is a good choice. If you have people with Unix backgrounds then I think FreeBSD or OpenBSD is better. And for many ASP systems Sun is the clear winner. However keep watch on Apple if IBM manages to restore hardware parity for Apple. I think that as a server OSX will mature quickly.
Calling this Carbon is a tad misleading. It isn't the same as the Carbon on OSX in terms of functionality. It might be better to call it a limited subset of Mac API calls.
Don't floppies deteriorate fairly quickly. (Relatively speaking) I know that back when floppies were common I had a *lot* that started getting data errors after only 2 years of storage. For that reason I usually made up multiple disks of backups.
My feeling is that if you have some old media with data you need off them you likely are in a whole world of hurt.
Realistically companies should have hired a few summer interns about 10 years ago to go through all those old backups and start copying them to tape. Now days with HDs as cheap as they are, a lot of older systems should have been backed up with the appropriate emulator for whatever platform. (Z-80, Apple][, CPM, etc.)
The question is who has data in those formats that still needs them? More than likely it is mainly scientific facilities with lots of recording data. I've heard that this is a *huge* problem in astronomy where a lot of old magnetic media from the 70's has decayed such that a lot of analysis has been lost.
Once again I'm just going by how long between 10.0 and when all the Apps started coming out from the major players. Several key applications still haven't come out in OSX and many applications have only come out the past couple of months. And that was with those companies having betas long before 10.0 came out.
With the switch to a new processor you can expect many more things to break. Apple undoubtedly can recompile all their iApps. The issue is third party companies. (Especially MS Office)
I think Apple has been hedging their bets because of problems with Motorola. Thus they've been compiling much of Darwin on both systems and likely Aqua as well. But to run them on x86 will undoubtedly create all sorts of problems. Not to mention optimization problems.
It might be nice in the long term, but consider what the initial user experience will be those first six months to one year. . .
Further most companies will be forced to develop for both PowerPC and x86 version of OSX.
Unless IBM falls behind schedule I just can't see Apple being able to afford the costs of switching to the x86. The real problems of the PowerPC chip in terms of competitive speed though means Apple has to be able to if necessary.
This is actually something I don't understand. On the one hand the 970 is single processor. Yet at the same time, as the Ars article explains, IBM clearly intends it for multi-processor systems. (i.e. like the dual 867 G4e that is popular at the moment)
If this will likely appear in multiprocessor systems with 2-4 processors, why make each die single processor?
I'm sure there is a reason - likely heat. Further Apple clearly will be selling single processor systems. (i.e. iBooks and iMacs) However it seems like there may be advantages to a dual processor on a single die as opposed to using the external bus to communicating. Perhaps this will be the next version of the 970?
It will never get released though unless both Motorola and IBM drop the ball. The reason is that x86 chips can not easily emulate PowerPC code. (The reverse is not true, due to the different ways the chips are designed) That would mean that if Apple moved to a x86 architecture you'd have BIG problems with backwards compatibility.
Apple's done some amazing switches. The 68040 to PowerPC went nearly flawlessly and amazed most people. Even Classic emulation on OSX is fairly impressive. But trying to emulate Photoshop on x86 would be nasty. Further look how long it took the big players (Macromedia, Adobe, etc.) to come out with OSX native programs. The switch to x86 would likely involve a two year wait for several important native apps. Not something Apple could withstand easily.
As for why Apple wouldn't go for the Power4, it is because it simply is too expensive. Further Apple wants something like the Altivect unit that is in current G4s. Power4 is simply optimized for non-desktop uses and is overkill for what Apple needs.
The 970 is a nice balance between Apple's needs and the Power core. Further by moving to IBM Apple is able to get a far better provider than Motorola whose G5 has been missing in action for some time now.
Apple will almost certainly go with 970 and not switch to Intel for the following reasons:
1. It is difficult to emulate PowerPC with Intel (although the reverse isn't *that* difficult). Apple would need a PowerPC emulator so that all that software (including OSX software) isn't lost.
2. Apple wants to differentiate itself somewhat from the PC.
3. IBM appears to be moving up after the several years of problems with Motorola.
The downside is that by the time a 970 board is out it will definitely be in the middle of the pack relative to the PC world. That means that Apple still will have computers that are more expensive than the PC world and that aren't as fast.
Of course I think OSX is sufficiently better than either Linux or WinXP for a workstation that I'll stick with it. But Apple best hope that IBM gets large yields on time and perhaps with better speeds than expected.
If this was back in the OSX 1.0 days you are right. With 10.2 it comes standard with a full Python distro along with the latest GNU tools. Further Fink has pretty much every tool compiled and debugged and installs them for you. This inlcudes X11 apps which admittedly once were a pain to run on OSX.
So this isn't a problem anymore and hasn't been for quite some time.
I once had a debate with one of the programmers here who was definitely a Unix geek. (Although he too has converted to OSX) I was complaining about my frustration in using Gnome everytime I had to work on a port of our software. The problem is that most Unix geeks are CLI only. I mean half the people I know don't even use a GUI for their debugging with gdb! And they love things like vi or the like for ALL their editing. (Most Emacs is better, but is *so* complex to configure that unless you cut your teeth on it your are better off not trying)
He had a point that being able to telnet or now SSL into a server and use these tools is faster. But that is but one fraction of the time and only for some users. For the rest I've never figured out why these "unix geeks" do things that are often the "hard way." Yeah they can do things faster, but only because they stick to what they already know and never learn other ways. Further, as you said, they don't do things like Word Processing or the like. So when they use X11 apps they think they actually are as good as equivalent Windows or Mac apps.
For the rest of us it sucks. The nice thing about OSX is that you have both available. It comes standard with an awesome from end for gcc and gdb. (Other than not saving watch variables between debugging sessions - damn that bugs me) You also have MS-Office and many other Apps that are awsome. (And yes I've tried OpenOffice - it just isn't as good IMO)
That's why I love OSX. It truly is the best of all worlds. Further it accomplished everything on the GUI/desktop that Linux folks have been saying they'd do. And the Linux folks have been saying that for years!
Don't get me wrong. There are still plenty of rough edges. Alias SymLink issues still pop up at odd times. The hardware is too slow and overpriced. But beyond that you have one amazing product in my opinion. Further every edition of OSX gets better and better. I can't wait until their new file system is out.
As others said, the "links" that the Finder makes are Aliases, not SymLinks. Aliases are more powerful as they combine features of hard links and soft links. Think of it like having a hard link which, if it fails, defaults to a sym link. Further Aliases contain information for connecting to other networks or the like if necessary. I really think Aliases are better than Symlinks, which are only one step better than Win shortcuts.
As I mentioned elsewhere there are scripts to convert Aliases to Symlinks for those places that need them. (Mainly directories seen by FTP and Apache) For everywhere else I think Aliases are superior and should be used.
If you want a GUI for doing true sym links download Path Finder. It is a great Finder replacement with quite a few features that most Unix users will love.
As for hidden files, I like the way Apple does that as well. It is nice to be able to hide files in more ways than simply prepending it with a period. i.e. hiding the Unix directories from regular users. If Linux ever wants to be a safe desktop OS it'll have to do this as well.
I don't think this is so. I didn't keep the application as I decided I had no need for a virtual desktop. (It was up there with labels for the "gee everyone gripes about this but I never used it when i had it" award) I do recall that during the week I tried it that I drug Finder windows between desktops fairly regularly.
I don't think so. If I recall it shipped with 3.1 and not 3.2. (Unless that change took place there as well) I don't know why some programs broke. If they depended on files being in certain locations or on undocumented APIs that were in flux I could understand it. But a few programs that broke made no sense.
Most Unixes do not simply use/etc/passwd or so forth. OpenBSD does not, for instance. Its been a while since I worked on it, but I believe Solaris 8.0 doesn't as well. So that is a rather common feature of most Unixes.
However for many other features Apple has been making things more standard.
Having said that I didn't say that NetInfo wasn't used. It is. I think it is used more for OSX system applications. (i.e. at the GUI level) So, for instance, most "hacks" for OSX dealing with UI elements involve NetInfo or related applications. However for the type of customizations that typical Unix people do, Apple has been moving to making their traditional hacks work. This, to me, is the best of both worlds. For one it means that many "how to" articles on the Internet for traditional Unix apps work. But you still have the elegance of NetInfo for features that are only available on OSX.
In other words if it is a function that is available on other Unixes, Apple will try and do it the traditional way. If it is a mainly Apple-only feature they'll do it the elegant way.
There are exceptions, such as users. But those are places where the flavors of Unix differ so much that it is difficult to say that the Linux way is the right way. (And frequently there are optional add-ons in these areas even for Linux)
I think the 255 char limit is gone. At least I don't recall encountering it in 10.2.
One "gotcha" I encountered though was trying to access HFS+ features from Python. The command line version of Python has the macfs module, but it appears to have been written for the old MacPython application. It assumes HFS paths and not Posix paths and then won't run half the time.
Of course this is more a Python problem due to the fact that under OS9 Python was an application. So hopefully the Python folks will fix this with time.
Bash is not the default shell. I don't even remember if Apple installs bash. I know you can get it with Fink. The default shell tcsh (which I like better anyway).
Slashdot Mirror
The problem is that most reasonably sized departments need an IT staff anyway. Having them run a mail server or the like isn't that big a deal. While some things can be effecient to subcontract out (i.e. your web server) often it is easier to have it on sight.
There are exceptions. But I think that only a few IT functions really can reasonably be marketed out. I think IBM's marketing strategy will work - but only for a small niche.
However for any application that is being accessed heavily and consistently it would seem that a Linix cluster or Sun cluster would be much more cost effective.
Those who, for example, might need a rendering farm but only for a short time might benefit. Consider that you only pay for the processing you need. If IBM comes up wiht some clustering software that is good and distributed this might work. However it is clearly aimed at the markets that are already buying very large IBM computers. It won't help, for instance, for the typical internet sever.
Having said that though, what kinds of people are that? The main rendering farms are being used fairly consistently. So for them having a bunch of Suns or equivalent systems is more efficient. They can then just add computers. So who is it that would need this sort of thing?
And if they did try and foist it on the general public it would obviously fail immediately. After all the heaviest users of processing time on general computers are games. And most people aren't quite willing to pay for the processing the latest Halo or equivalent might use. (Not to mention the fact that Dell, Apple, and Compact wouldn't follow IBM)
There is always Python. Wasn't one of the larger Yahoo-like services done in Python? I can't recall which one though.
There are problems with this, not the least of which are the silly benchmarks the author used and the apparent lack of optimizations. Further anyone running Linux will either be running it on faster x86 hardware or is running it on a Mac solely because they have old hardware not fast enough to run OSX.
So you are right. Even beyond the benchmark problems it ends up being of limited relevance for deciding on a server.
Yeah most people here would rather roll their own hardware/software system with either Linux or BSD on an x86 system. For larger systems Sun is the reasonable choice. But just because it doesn't meet every need isn't to say that it can't be successful for a server. After al there are different kinds of servers. You wouldn't compare a powerful multiprocessor Sun against Linux on a Pentium4, for instance. They simply are focused in on different markets.
Getting back to your hardware point. Other than focusing in on obsolete hardware, you do have a point. Right now the G4 and chipset Apple uses isn't competitive. Its amazing they got the performance/dollar they did. However with the 970 next year Apple should be able to move into being far more competitive.
Yeah there is always the danger of "vaporware" comparisons. (Remember when the G5 was just around the corner?) But I think IBM is far more reliable than Motorola. Further IBM likely wants to use the chip in their own Linux multiprocessor systems. In that case we'll have a very interesting comparison between OSX and Linux. Further I suspect that while the 970 will initially be middle of the pack, that it may end up doing quite well in multiprocessor servers. By that time OSX and their server software will be in their next major revision and should be far more mature. Hopefully the new filesystem will even be out.
Of course if you really need full remote control there are all sorts of programs like that for both XP and OSX. Of course they are much slower than X11 whether running on OSX or XP) But then you typically don't need that functionality much. Further those using them probably are running across a 1GB Ethernet communication rather than across the Internet.
For more info check out Apple's XServe site
or do a google for some reviews on the GUI. But if you are going to make a claim that there is no remote GUI management is simply silly.In those cases the ease of management will likely be a bigger factor than overall speed. If overall speed is that important you really need someone who is an expert in Linux or BSD, knows how to eck out every bit of speed, knows security up the yazoo, etc. And of course your department will have to pay for this guy. And that ends up being near $100,000 a year once you start including benefits etc. (Or more if you have several folks)
For a more regular IT staff who doesn't want to spend that much time or effort XServe, like the equivalent offerings from DELL, Sun or Compact, is a nice server. As I said for ASP I still think Sun might be a better choice. But it really depends upon what you are doing.
One thing to remember is that OSX is still using the HFS+ file system. The Byte article didn't mention is they were comparing it with the UFS. In either case Apple has new file systems coming down that might shake things up a bit in terms of both reliability and speed. Of course Linux has those now. But I think everyone agrees that XServe is nice for a first generation product. However as with most first generation products, think twice about being on the cutting edge. A better comparison might be a year from now when OSX is up to 10.3 with the equivalent XServe. Hopefully by then the new hardware based on the 970 will be out. Of course by the same measure the Linux/x86 world will also have moved on.
Xinet Benchmark
I admit I'm a tad skeptical of the relevance of this benchmark, but it does seem that Apple has a nice system. I suspect you could roll your own better with OpenBSD or Linux and a nice AMD multiprocessor system. That's just me though. And realistically a lot of businesses DON'T want such systems. They want a "come as it is" system. Further a lot of people don't want all the messing around that you have to do with most Linux of BSD distributions. Apple has put a very nice interface on their server. Yet you have the added benefit of being able to drop to Unix when necessary.
Apple's big problem is still the chipset used with the G4's. Given that, despite many of the nice features, unless you are primarily serving other Macs, I don't think XServe is a good choice. If you have people with Unix backgrounds then I think FreeBSD or OpenBSD is better. And for many ASP systems Sun is the clear winner. However keep watch on Apple if IBM manages to restore hardware parity for Apple. I think that as a server OSX will mature quickly.
Calling this Carbon is a tad misleading. It isn't the same as the Carbon on OSX in terms of functionality. It might be better to call it a limited subset of Mac API calls.
My feeling is that if you have some old media with data you need off them you likely are in a whole world of hurt.
Realistically companies should have hired a few summer interns about 10 years ago to go through all those old backups and start copying them to tape. Now days with HDs as cheap as they are, a lot of older systems should have been backed up with the appropriate emulator for whatever platform. (Z-80, Apple][, CPM, etc.)
The question is who has data in those formats that still needs them? More than likely it is mainly scientific facilities with lots of recording data. I've heard that this is a *huge* problem in astronomy where a lot of old magnetic media from the 70's has decayed such that a lot of analysis has been lost.
With the switch to a new processor you can expect many more things to break. Apple undoubtedly can recompile all their iApps. The issue is third party companies. (Especially MS Office)
I think Apple has been hedging their bets because of problems with Motorola. Thus they've been compiling much of Darwin on both systems and likely Aqua as well. But to run them on x86 will undoubtedly create all sorts of problems. Not to mention optimization problems.
It might be nice in the long term, but consider what the initial user experience will be those first six months to one year. . .
Further most companies will be forced to develop for both PowerPC and x86 version of OSX.
Unless IBM falls behind schedule I just can't see Apple being able to afford the costs of switching to the x86. The real problems of the PowerPC chip in terms of competitive speed though means Apple has to be able to if necessary.
If this will likely appear in multiprocessor systems with 2-4 processors, why make each die single processor?
I'm sure there is a reason - likely heat. Further Apple clearly will be selling single processor systems. (i.e. iBooks and iMacs) However it seems like there may be advantages to a dual processor on a single die as opposed to using the external bus to communicating. Perhaps this will be the next version of the 970?
Apple's done some amazing switches. The 68040 to PowerPC went nearly flawlessly and amazed most people. Even Classic emulation on OSX is fairly impressive. But trying to emulate Photoshop on x86 would be nasty. Further look how long it took the big players (Macromedia, Adobe, etc.) to come out with OSX native programs. The switch to x86 would likely involve a two year wait for several important native apps. Not something Apple could withstand easily.
As for why Apple wouldn't go for the Power4, it is because it simply is too expensive. Further Apple wants something like the Altivect unit that is in current G4s. Power4 is simply optimized for non-desktop uses and is overkill for what Apple needs. The 970 is a nice balance between Apple's needs and the Power core. Further by moving to IBM Apple is able to get a far better provider than Motorola whose G5 has been missing in action for some time now.
Apple will almost certainly go with 970 and not switch to Intel for the following reasons: 1. It is difficult to emulate PowerPC with Intel (although the reverse isn't *that* difficult). Apple would need a PowerPC emulator so that all that software (including OSX software) isn't lost. 2. Apple wants to differentiate itself somewhat from the PC. 3. IBM appears to be moving up after the several years of problems with Motorola. The downside is that by the time a 970 board is out it will definitely be in the middle of the pack relative to the PC world. That means that Apple still will have computers that are more expensive than the PC world and that aren't as fast. Of course I think OSX is sufficiently better than either Linux or WinXP for a workstation that I'll stick with it. But Apple best hope that IBM gets large yields on time and perhaps with better speeds than expected.
So this isn't a problem anymore and hasn't been for quite some time.
He had a point that being able to telnet or now SSL into a server and use these tools is faster. But that is but one fraction of the time and only for some users. For the rest I've never figured out why these "unix geeks" do things that are often the "hard way." Yeah they can do things faster, but only because they stick to what they already know and never learn other ways. Further, as you said, they don't do things like Word Processing or the like. So when they use X11 apps they think they actually are as good as equivalent Windows or Mac apps.
For the rest of us it sucks. The nice thing about OSX is that you have both available. It comes standard with an awesome from end for gcc and gdb. (Other than not saving watch variables between debugging sessions - damn that bugs me) You also have MS-Office and many other Apps that are awsome. (And yes I've tried OpenOffice - it just isn't as good IMO)
That's why I love OSX. It truly is the best of all worlds. Further it accomplished everything on the GUI/desktop that Linux folks have been saying they'd do. And the Linux folks have been saying that for years!
Don't get me wrong. There are still plenty of rough edges. Alias SymLink issues still pop up at odd times. The hardware is too slow and overpriced. But beyond that you have one amazing product in my opinion. Further every edition of OSX gets better and better. I can't wait until their new file system is out.
As I mentioned elsewhere there are scripts to convert Aliases to Symlinks for those places that need them. (Mainly directories seen by FTP and Apache) For everywhere else I think Aliases are superior and should be used.
If you want a GUI for doing true sym links download Path Finder. It is a great Finder replacement with quite a few features that most Unix users will love.
As for hidden files, I like the way Apple does that as well. It is nice to be able to hide files in more ways than simply prepending it with a period. i.e. hiding the Unix directories from regular users. If Linux ever wants to be a safe desktop OS it'll have to do this as well.
I don't think this is so. I didn't keep the application as I decided I had no need for a virtual desktop. (It was up there with labels for the "gee everyone gripes about this but I never used it when i had it" award) I do recall that during the week I tried it that I drug Finder windows between desktops fairly regularly.
I don't think so. If I recall it shipped with 3.1 and not 3.2. (Unless that change took place there as well) I don't know why some programs broke. If they depended on files being in certain locations or on undocumented APIs that were in flux I could understand it. But a few programs that broke made no sense.
However for many other features Apple has been making things more standard.
Having said that I didn't say that NetInfo wasn't used. It is. I think it is used more for OSX system applications. (i.e. at the GUI level) So, for instance, most "hacks" for OSX dealing with UI elements involve NetInfo or related applications. However for the type of customizations that typical Unix people do, Apple has been moving to making their traditional hacks work. This, to me, is the best of both worlds. For one it means that many "how to" articles on the Internet for traditional Unix apps work. But you still have the elegance of NetInfo for features that are only available on OSX.
In other words if it is a function that is available on other Unixes, Apple will try and do it the traditional way. If it is a mainly Apple-only feature they'll do it the elegant way.
There are exceptions, such as users. But those are places where the flavors of Unix differ so much that it is difficult to say that the Linux way is the right way. (And frequently there are optional add-ons in these areas even for Linux)
One "gotcha" I encountered though was trying to access HFS+ features from Python. The command line version of Python has the macfs module, but it appears to have been written for the old MacPython application. It assumes HFS paths and not Posix paths and then won't run half the time.
Of course this is more a Python problem due to the fact that under OS9 Python was an application. So hopefully the Python folks will fix this with time.
Bash is not the default shell. I don't even remember if Apple installs bash. I know you can get it with Fink. The default shell tcsh (which I like better anyway).