You're confusing the iTunes Music Store with regular iTunes files.
I've got my iTunes files also served by a MediaTomb DLNA server to everything that isn't Apple in the house: PlayStation 3, XBMC, streaming app on the iPod Touch, and so on. I've used SoundJam MP and its successor, iTunes, to rip and encode MP3 files since I got a machine fast enough to play MP3 instead of MP2.
Those files all play back perfectly on an el-cheapo Chinese no-name MP3 CD player, the Rio Volt MP3 CD player, the Sony headunit I put in my old car, the Subaru factory stereo in my current car, my stepfather's yay-company-promotional flash-drive MP3 player, my Garmin Zumo GPS's built-in MP3 player, and so on.
Apps are, of course, the new lock-in--just like my Linux apps don't run on OS X.
It's a good time to test your full DR plan: see what it takes to bring up your entire production needs from backups.
Obviously, environments with lots of load-balancing servers would want to just bring up one or two, not the full capacity.
And yes, I've sat with the tape drives, the DR plan, and a red pen going through the procedure and re-writing it until it actually worked.
From then on, any time a disk failed, it was a 5 minute service call. "You've got backups?" "Yup." "Great. These aren't supposed to be hot-swapped, so please turn around while I do something here..."
While I agree with out about Netscape 4 in every possible way--it could be made stable-ish on UNIX and Mac by turning off everything that wasn't in 3....
To me, the point of the anti-competition claim wasn't that Netscape got screwed.
It's much more about all the projects that never got started, or never saw the light of day, because people looked at what happened to Netscape and didn't even try.
That's the important thing.
But at least, on the Mac, we had iCab for our happy browsing needs.
Not all of us buy pre-built PCs. Motherboards and cases do not come with Windows licenses.
I was also able to find a little Zotac Atom machine that came with no operating system (or disk drive to put it on, or RAM to run it in). With the addition of a bit of RAM, it runs Linux just fine over the network.
Anyway, OEM licenses are not transferable: he'd have to be running WINE on the machine that came with the license.
Kinect is evolutionary; you'll want to have seen the Mandala system on an Amiga 1000 in 1988 or 1989. The ideas have been getting worked on in a number of places for years.
I'll accept an argument that Kinect's price point is revolutionary. Most people would charge what it actually cost to develop and manufacture, which would keep it out of the consumer market.
You can also use this to make your own password-protected sparsebundles. When I do stuff on contract, I use an encrypted sparsebundle for that client. Double click on the sparsebundle, and it prompts for the password and mounts.
I do keep symlinks to the/Volumes/BundleName directories in $HOME for command line convenience. But when the bundle isn't mounted, all tbe links are dead and point somewhere I can't write, so I can't accidentally create files.
Get started with "New Disk Image" in Disk Utility.
It still doesn't matter if Google is including kernel-specific definitions. As long as they are interface definitions, they cannot be protected by copyright, as they are not creative works. They are simply statements of facts, like:
void snd_emu8000_ops_setup(emu8000_t *emu);
That's not copyrightable, and it's not useful in user-space. (It's from code not even compiled on my version of RHEL.) Putting a copyright notice on the file doesn't make factual information copyright; it indicates who holds the copyright on copyrightable material in the file. (Just about every country involved is a Berne convention signatory these days, so copyright always adheres to creative work, even without a notice.)
So, as long as Google's mechanical transform retains only function declarations and no definitions (per ISO C use of the words), it's fine. Structure and union definitions, on the other hand, are required and would also be allowed--as they are simply statements of facts about how information should be organized to be fed to a function.
But Linus's point is, if Google included stuff from the kernel-only section, it's useless to userland code. All they have to do is remove it. It won't affect anything in userland because they couldn't use it anyway, the definitions only work for something being loaded as a kernel module or compiled right in to the kernel.
That's why he's not really worried about it. It's either stuff for the userland-kernel interface, which is expressly licensed to avoid having GPL adhere, or it's stuff that can't be used from userland, which means the "harm" can be trivially resolved by stopping distribution of those files--assuming copyrightable material was included in the first place. Which is a big assume for C header files; even with inline functions and macros.
While I'm in no way a fan of C++ templates, or even the whole language, and you are technically correct (the best kind of correct)....
For templates to be truly interesting, they must work with user-defined types. And that means you need to be able to create new specializations, ones the original author/vendor of the template library didn't think of. To create a specialization, the implementation must be available along with the new type--which means the library code and the user code, together in one translation unit.
If you work around that by using pointers or require everything inherit from a specific base, there's probably not much reason to have used templates after all.
If it was truly usage-based, there'd be no cap. You'd start with a basic "0 byte" connection for whatever it costs to operate the line, then pay per GB. (Why 40 GB increments? Why not 128MB or something reasonable? I don't buy electricity in 100 kWh blocks; my meter runs to the 1/10th of a kWh. Sure, advertise the rate as $X per 40 GB if you like, but bill fractionally.)
Thing is, the dominant cost of the network is the static, "0 byte" service. The incremental cost of transfer is very small compared to the cost of bandwidth provisioning in the first place. The billing system alone could cost more than the transfer costs.
And the real problem was, of course, forcing a usage-based billing model "because we can't compete with them." 3rd party ISPs already pay for bandwidth; they have to lease aggregation lines back to (say) 151 Front Street, they have to pay for peering to get on the Real Internet, their customers pay for the line capacity from the DSLAM to their house... all of that is already being paid for on a capacity basis. (More customers for ISP X? ISP X has to buy more aggregation and have a larger peering agreement.)
So it wasn't just the price, which was insanely out of whack with real costs. It was the double-dip. It's one of the best examples of "Regulatory Capture enabling Rent Seeking".
Ugh, the pre-fluid dynamic bearing Maxtors.... They'd be so sweet and quiet when you got them home from the shop, you'd install them and be very pleased with your purchase.
Then, 3 weeks to 3 months in... that buzzing bearing sound would start. Quiet, at first... and intermittent. Just enough so that you weren't really sure you'd heard anything at all. But gradually, it would get louder... and louder... and louder... until there was no debate, the affair was over: the 60giger was screaming louder than a cat left out in the rain. And with much the same sound.
That drive _still worked_ when we pulled it from a friend's system and zeroed it for disposal 2 weekends ago. But damn, did you know when that system was powered on.
Even some speakers are assembled this way; I've got a pair of Logitech V20 USB-powered speakers that failed. Thinking, "I'm not bad with a soldering iron and test probe", I took out the 4 screws... that merely serve to hold the case together while the adhesive sets. 5 minutes with a spudger finally cracked the case open.
(Once inside, I could see why they glued it shut--well, aside from the obvious "buzz" prevention--the PCB manufacturing is horrible. Really, really bad: uneven layering of the solder mask I could forgive, but all of the through-hole soldering could serve as a textbook photo of a "cold solder joint". And there's either flux residue or something worse on the power filter cap. It reminds me of my soldering... from when I was 10 and had no idea what I was doing.)
Well, I don't trust the prompting for account name and password. There's absolutely nothing that confirms, to me, that the prompt is from Apple's code and not the game.
Writing a fake login program was one of the first things we played with back in 1985 in high school.
Isn't that exactly like the "wanna-f**k" sites and apps that tell you someone is interested in you only if you are interested in them?
That is, if I say I'm interested in Person X, nothing happens. But when Person X also says they're interested in me, we both then get some indication that has happened--instant message, profile marker, whatever. It's really great for shy people. Though you still have the "I don't know what to say" issue.
This one goes back to multi-user BBS days....
Come to think of it, I might have seen something similar, like "don't list extreme fetish interests for users who have not filled out any extreme fetish interests". Hmmm. I wonder if those ex-sysops I know have backups of the BBS software....
Such as system is in use in Toronto, Ontario, Canada. Not for data centres specifically, but for air conditioning of downtown office buildings.
It's called Deep Lake Water Cooling, and pulls 4degC water in from the deeper parts of the lake. The Enwave cooling plant then warms it up just a little bit, sinking heat from the buildings on its central cooling service. From there, the water goes to the treatment plants for the City's domestic water supply.
From a citizen's perspective, tap water in late summer is cooler than it was before the system was built; the old--shallower--intakes were affected by lake heating in the summer. To the point where, in some years, by August the tap water was a bit manky because of the algae levels and corresponding increase in treatment chemicals. Don't seem to get that any more.
For the City, an independent company built a whole new intake and pumping system, saving the works department the expense--the old intakes were on their last legs anyway.
Obviously, part of why it works is because only one City on the lake is doing it.
Toronto is not known for "dry heat" in summer. "Kill me now before the humidity does" is more like it... though Hong Kong and Washington, D.C. have us well-beat there, we do like to complain about the weather.
Sega Genesis: Motorola 68000 NES: Ricoh 2A03 (MOS 6502 core) Atari 2600: MOS 6507 (MOS 6502 with only 8K addressing)
Sure, the video and sound hardware might have been somewhat "specialized". But contemporary personal computers had "specialized" sound and graphics. The VIC-20 was different from the Apple ][ was different from the Atari 400 was different from....
CPUs were pretty much always off-the-shelf. They just weren't off the IBM PC(r) shelf. Another example; the Hitachi SuperH CPUs of the Sega Saturn and 32X are also used in control systems.
I miss RHL 6.2. That was as stable and clean an OS as I've used.
Want the ISOs? I've still got 'em (somewhere...). And 7.3. And 9. And Fedora Core 3 and 4. Source and all.
That's the thing about Linux: you can keep it. There's no "activation" service which is going to go away. I've been backporting kernel patches into my Fedora 11 server because upgrading is more painful. (Though I do plan to move it to CentOS 6 when it [finally] comes out; I should have just used CentOS in the first place--would have saved time removing desktop stuff and adding server stuff.)
But I do like my 8-channel SAS card that can manage 800 MB/s sustained transfer on an MD stripe set. So I don't miss the second-hand 486/66 with a SCSI-II Fast controller. (Woohoo! 10 MB/s!) I really don't miss how it needed to boot from floppy or IDE HDD only....
I did pick up a 386/25 last night. I haven't decided if it's going to be taken apart for the gates and support chips and motors and magnets... or if it's going to be a serial terminal for fiddling with microcontrollers. How does VGA work, again?
If you want $1 DVDs, try Chinatown Centre in Toronto, on Spadina between Queen and Dundas. Or Pacific Mall in Markham, on Steeles near Kennedy (IIRC).
Now, most of the $1 DVDs are pirated Chinese DVDs. But there's Hollywood stuff in there, too; just usually not shown through the window.
They get busted every year or so, but a month later the shops are full again. For Chinatown Centre, no-one else seems to want to rent those spaces; it's quite the disaster of a mall in terms of footage rented out. So the property managers aren't going to try to stop things. The customers sure aren't.
And if their products are so unreliable that an extended warranty is a good idea, why do I want to buy it in the first place?
Saying that, turning, and walking out of the store is extremely fun, BTW.
To be honest, though, I got it from my grandmother: She was buying a washer, and the salesman was going through all the "Maytag is so reliable" things (back when they still were). Then, after writing up the sale, he started in on the extended warranty. My grandmother just said, "You just told me how reliable these machines are. Now you're trying to sell me a warranty? Forget it, I'll go somewhere else and get a different one." There was quite a lot of backpedalling and a Maytag did eventually turn up at her apartment, sans extra warranty.
It's a bootstrap thing. Most Arduino kits--you might want to check out the Solarbotics Freeduino-SB--come with a microcontroller pre-programmed with a bootloader.
Once you've got that, you can use the Arduino to drive the in-circuit serial programming interface to program another MCU. Otherwise, you'll need some sort of ICSP interface. (If you've got one anyway, you're done, stop reading.)
Most USB-based Arduinos (and clones) can also drive the ICSP from the USB-async bridge chip. The Freeduino-SB has pin-headers making this easy, but all devices with the same USB bridge can do it. This lets you drive the ICSP from your PC (OSX , Linux, or Windows); the target MCU doesn't even have to be on the Arduino board.
So, it's a good way to bootstrap your first MCU. And the I/O shields are kind of fun, too... but there's a phenomenal number of LEDs on the LOLshield... it doesn't really sink in until you actually sit down to put it together....
(No relation to Solarbotics other than we're both Canadian. It just happens to be the Freeduino I got to deal with said bootstrapping.)
I was hired at my first job in the field I was studying much more on the basis of what I'd done with my hobbies than what I'd heard in the lecture hall. Mind you, at the time my hobbies included writing a buffered I/O library in 68000 assembly.
In business, I've found the people who don't "play" with computers or technology also aren't very good at "working" with it.
Basically, it's a way to separate the people who are actually interested in the work from those who are looking for a clock to punch and a paycheque to cash.
Slashdot Mirror
Apple's OTHER secret weapon is:
They are not trying to be, "All things to all people."
They are trying to be, "This thing to people that want it."
You want to do other things? You want another product. They're actually fine with that.
You're confusing the iTunes Music Store with regular iTunes files.
I've got my iTunes files also served by a MediaTomb DLNA server to everything that isn't Apple in the house: PlayStation 3, XBMC, streaming app on the iPod Touch, and so on. I've used SoundJam MP and its successor, iTunes, to rip and encode MP3 files since I got a machine fast enough to play MP3 instead of MP2.
Those files all play back perfectly on an el-cheapo Chinese no-name MP3 CD player, the Rio Volt MP3 CD player, the Sony headunit I put in my old car, the Subaru factory stereo in my current car, my stepfather's yay-company-promotional flash-drive MP3 player, my Garmin Zumo GPS's built-in MP3 player, and so on.
Apps are, of course, the new lock-in--just like my Linux apps don't run on OS X.
It's a good time to test your full DR plan: see what it takes to bring up your entire production needs from backups.
Obviously, environments with lots of load-balancing servers would want to just bring up one or two, not the full capacity.
And yes, I've sat with the tape drives, the DR plan, and a red pen going through the procedure and re-writing it until it actually worked.
From then on, any time a disk failed, it was a 5 minute service call. "You've got backups?" "Yup." "Great. These aren't supposed to be hot-swapped, so please turn around while I do something here..."
While I agree with out about Netscape 4 in every possible way--it could be made stable-ish on UNIX and Mac by turning off everything that wasn't in 3....
To me, the point of the anti-competition claim wasn't that Netscape got screwed.
It's much more about all the projects that never got started, or never saw the light of day, because people looked at what happened to Netscape and didn't even try.
That's the important thing.
But at least, on the Mac, we had iCab for our happy browsing needs.
Not all of us buy pre-built PCs. Motherboards and cases do not come with Windows licenses.
I was also able to find a little Zotac Atom machine that came with no operating system (or disk drive to put it on, or RAM to run it in). With the addition of a bit of RAM, it runs Linux just fine over the network.
Anyway, OEM licenses are not transferable: he'd have to be running WINE on the machine that came with the license.
Kinect is evolutionary; you'll want to have seen the Mandala system on an Amiga 1000 in 1988 or 1989. The ideas have been getting worked on in a number of places for years.
I'll accept an argument that Kinect's price point is revolutionary. Most people would charge what it actually cost to develop and manufacture, which would keep it out of the consumer market.
You're confusing "benefit Canada" with "benefit the politicians who approve the purchase".
OK, so sure, the former is the law and the latter is what would actually happen.
You can also use this to make your own password-protected sparsebundles. When I do stuff on contract, I use an encrypted sparsebundle for that client. Double click on the sparsebundle, and it prompts for the password and mounts.
I do keep symlinks to the /Volumes/BundleName directories in $HOME for command line convenience. But when the bundle isn't mounted, all tbe links are dead and point somewhere I can't write, so I can't accidentally create files.
Get started with "New Disk Image" in Disk Utility.
You might also have a network that's got just enough IPv6 working that FF tries AAAA records before A records.
If it's still there in 4, try frobbing network.dns.disableIPv6 in about:config to see if that makes a difference.
It still doesn't matter if Google is including kernel-specific definitions. As long as they are interface definitions, they cannot be protected by copyright, as they are not creative works. They are simply statements of facts, like:
That's not copyrightable, and it's not useful in user-space. (It's from code not even compiled on my version of RHEL.) Putting a copyright notice on the file doesn't make factual information copyright; it indicates who holds the copyright on copyrightable material in the file. (Just about every country involved is a Berne convention signatory these days, so copyright always adheres to creative work, even without a notice.)
So, as long as Google's mechanical transform retains only function declarations and no definitions (per ISO C use of the words), it's fine. Structure and union definitions, on the other hand, are required and would also be allowed--as they are simply statements of facts about how information should be organized to be fed to a function.
But Linus's point is, if Google included stuff from the kernel-only section, it's useless to userland code. All they have to do is remove it. It won't affect anything in userland because they couldn't use it anyway, the definitions only work for something being loaded as a kernel module or compiled right in to the kernel.
That's why he's not really worried about it. It's either stuff for the userland-kernel interface, which is expressly licensed to avoid having GPL adhere, or it's stuff that can't be used from userland, which means the "harm" can be trivially resolved by stopping distribution of those files--assuming copyrightable material was included in the first place. Which is a big assume for C header files; even with inline functions and macros.
While I'm in no way a fan of C++ templates, or even the whole language, and you are technically correct (the best kind of correct)....
For templates to be truly interesting, they must work with user-defined types. And that means you need to be able to create new specializations, ones the original author/vendor of the template library didn't think of. To create a specialization, the implementation must be available along with the new type--which means the library code and the user code, together in one translation unit.
If you work around that by using pointers or require everything inherit from a specific base, there's probably not much reason to have used templates after all.
If it was truly usage-based, there'd be no cap. You'd start with a basic "0 byte" connection for whatever it costs to operate the line, then pay per GB. (Why 40 GB increments? Why not 128MB or something reasonable? I don't buy electricity in 100 kWh blocks; my meter runs to the 1/10th of a kWh. Sure, advertise the rate as $X per 40 GB if you like, but bill fractionally.)
Thing is, the dominant cost of the network is the static, "0 byte" service. The incremental cost of transfer is very small compared to the cost of bandwidth provisioning in the first place. The billing system alone could cost more than the transfer costs.
And the real problem was, of course, forcing a usage-based billing model "because we can't compete with them." 3rd party ISPs already pay for bandwidth; they have to lease aggregation lines back to (say) 151 Front Street, they have to pay for peering to get on the Real Internet, their customers pay for the line capacity from the DSLAM to their house... all of that is already being paid for on a capacity basis. (More customers for ISP X? ISP X has to buy more aggregation and have a larger peering agreement.)
So it wasn't just the price, which was insanely out of whack with real costs. It was the double-dip. It's one of the best examples of "Regulatory Capture enabling Rent Seeking".
Bah, there's an app to remember passcodes.
You just have to remember the passcode for the app....
Ugh, the pre-fluid dynamic bearing Maxtors.... They'd be so sweet and quiet when you got them home from the shop, you'd install them and be very pleased with your purchase.
Then, 3 weeks to 3 months in... that buzzing bearing sound would start. Quiet, at first... and intermittent. Just enough so that you weren't really sure you'd heard anything at all. But gradually, it would get louder... and louder... and louder... until there was no debate, the affair was over: the 60giger was screaming louder than a cat left out in the rain. And with much the same sound.
That drive _still worked_ when we pulled it from a friend's system and zeroed it for disposal 2 weekends ago. But damn, did you know when that system was powered on.
Even some speakers are assembled this way; I've got a pair of Logitech V20 USB-powered speakers that failed. Thinking, "I'm not bad with a soldering iron and test probe", I took out the 4 screws... that merely serve to hold the case together while the adhesive sets. 5 minutes with a spudger finally cracked the case open.
(Once inside, I could see why they glued it shut--well, aside from the obvious "buzz" prevention--the PCB manufacturing is horrible. Really, really bad: uneven layering of the solder mask I could forgive, but all of the through-hole soldering could serve as a textbook photo of a "cold solder joint". And there's either flux residue or something worse on the power filter cap. It reminds me of my soldering... from when I was 10 and had no idea what I was doing.)
Well, I don't trust the prompting for account name and password. There's absolutely nothing that confirms, to me, that the prompt is from Apple's code and not the game.
Writing a fake login program was one of the first things we played with back in 1985 in high school.
But...
Isn't that exactly like the "wanna-f**k" sites and apps that tell you someone is interested in you only if you are interested in them?
That is, if I say I'm interested in Person X, nothing happens. But when Person X also says they're interested in me, we both then get some indication that has happened--instant message, profile marker, whatever. It's really great for shy people. Though you still have the "I don't know what to say" issue.
This one goes back to multi-user BBS days....
Come to think of it, I might have seen something similar, like "don't list extreme fetish interests for users who have not filled out any extreme fetish interests". Hmmm. I wonder if those ex-sysops I know have backups of the BBS software....
Such as system is in use in Toronto, Ontario, Canada. Not for data centres specifically, but for air conditioning of downtown office buildings.
It's called Deep Lake Water Cooling, and pulls 4degC water in from the deeper parts of the lake. The Enwave cooling plant then warms it up just a little bit, sinking heat from the buildings on its central cooling service. From there, the water goes to the treatment plants for the City's domestic water supply.
From a citizen's perspective, tap water in late summer is cooler than it was before the system was built; the old--shallower--intakes were affected by lake heating in the summer. To the point where, in some years, by August the tap water was a bit manky because of the algae levels and corresponding increase in treatment chemicals. Don't seem to get that any more.
For the City, an independent company built a whole new intake and pumping system, saving the works department the expense--the old intakes were on their last legs anyway.
Obviously, part of why it works is because only one City on the lake is doing it.
Toronto is not known for "dry heat" in summer. "Kill me now before the humidity does" is more like it... though Hong Kong and Washington, D.C. have us well-beat there, we do like to complain about the weather.
Hmmm.
Sega Genesis: Motorola 68000
NES: Ricoh 2A03 (MOS 6502 core)
Atari 2600: MOS 6507 (MOS 6502 with only 8K addressing)
Sure, the video and sound hardware might have been somewhat "specialized". But contemporary personal computers had "specialized" sound and graphics. The VIC-20 was different from the Apple ][ was different from the Atari 400 was different from....
CPUs were pretty much always off-the-shelf. They just weren't off the IBM PC(r) shelf. Another example; the Hitachi SuperH CPUs of the Sega Saturn and 32X are also used in control systems.
Want the ISOs? I've still got 'em (somewhere...). And 7.3. And 9. And Fedora Core 3 and 4. Source and all.
That's the thing about Linux: you can keep it. There's no "activation" service which is going to go away. I've been backporting kernel patches into my Fedora 11 server because upgrading is more painful. (Though I do plan to move it to CentOS 6 when it [finally] comes out; I should have just used CentOS in the first place--would have saved time removing desktop stuff and adding server stuff.)
But I do like my 8-channel SAS card that can manage 800 MB/s sustained transfer on an MD stripe set. So I don't miss the second-hand 486/66 with a SCSI-II Fast controller. (Woohoo! 10 MB/s!) I really don't miss how it needed to boot from floppy or IDE HDD only....
I did pick up a 386/25 last night. I haven't decided if it's going to be taken apart for the gates and support chips and motors and magnets... or if it's going to be a serial terminal for fiddling with microcontrollers. How does VGA work, again?
If you want $1 DVDs, try Chinatown Centre in Toronto, on Spadina between Queen and Dundas. Or Pacific Mall in Markham, on Steeles near Kennedy (IIRC).
Now, most of the $1 DVDs are pirated Chinese DVDs. But there's Hollywood stuff in there, too; just usually not shown through the window.
They get busted every year or so, but a month later the shops are full again. For Chinatown Centre, no-one else seems to want to rent those spaces; it's quite the disaster of a mall in terms of footage rented out. So the property managers aren't going to try to stop things. The customers sure aren't.
And if their products are so unreliable that an extended warranty is a good idea, why do I want to buy it in the first place?
Saying that, turning, and walking out of the store is extremely fun, BTW.
To be honest, though, I got it from my grandmother: She was buying a washer, and the salesman was going through all the "Maytag is so reliable" things (back when they still were). Then, after writing up the sale, he started in on the extended warranty. My grandmother just said, "You just told me how reliable these machines are. Now you're trying to sell me a warranty? Forget it, I'll go somewhere else and get a different one." There was quite a lot of backpedalling and a Maytag did eventually turn up at her apartment, sans extra warranty.
I see. Is it because of your mother that you say, "Hollywood is just an Emacs mode"?
It's a bootstrap thing. Most Arduino kits--you might want to check out the Solarbotics Freeduino-SB--come with a microcontroller pre-programmed with a bootloader.
Once you've got that, you can use the Arduino to drive the in-circuit serial programming interface to program another MCU. Otherwise, you'll need some sort of ICSP interface. (If you've got one anyway, you're done, stop reading.)
Most USB-based Arduinos (and clones) can also drive the ICSP from the USB-async bridge chip. The Freeduino-SB has pin-headers making this easy, but all devices with the same USB bridge can do it. This lets you drive the ICSP from your PC (OSX , Linux, or Windows); the target MCU doesn't even have to be on the Arduino board.
So, it's a good way to bootstrap your first MCU. And the I/O shields are kind of fun, too... but there's a phenomenal number of LEDs on the LOLshield... it doesn't really sink in until you actually sit down to put it together....
(No relation to Solarbotics other than we're both Canadian. It just happens to be the Freeduino I got to deal with said bootstrapping.)
I was hired at my first job in the field I was studying much more on the basis of what I'd done with my hobbies than what I'd heard in the lecture hall. Mind you, at the time my hobbies included writing a buffered I/O library in 68000 assembly.
In business, I've found the people who don't "play" with computers or technology also aren't very good at "working" with it.
Basically, it's a way to separate the people who are actually interested in the work from those who are looking for a clock to punch and a paycheque to cash.