I had the older version of their OTA DVR. It was sad when Rovi (the company that brought us Macrovision) suddenly shut down the OTA TV Guide service. It was great having two weeks of guide data. Now you're lucky to get 24 hours. Why? Because TV set manufacturers suck, and some older (and probably not-so-old) TV sets can't handle that much guide data, so stations don't send much.
I just checked, and all but three of my local stations have 24 hour guide data. Of the other three (at least 8 sub-channels total) one has 36 hours, and the other two have 72 hours of data.
Sure, I could subscribe to a feed for my MythTV, and reading the EPG data on my Hauppage cards sometimes hoses it such that I have a cron job to detect it and reboot, but free is free, dammit.
Some USB receivers have mini antennae, but I've never found them to work in any location, not even in the countryside or a downtown hotel.
I recently read that the reason they don't work well is that they need a ground plane. Notice how they have a really nice magnet? You need to stick that to a big piece of metal, like the roof of a car, for the antenna to work properly. Otherwise you might as well just jam a long piece of wire (like 2m-5m) into the connector and move it around until you get good reception.
Under US copyright law, the "strike" (image) of a font can not be copyrighted. Computer outline fonts can be copyrighted because the outlines are considered a computer program, even more so when there is hinting. Note that this copyright is on the specific outlines of a font. If you make a clone that looks the same, it will likely at least have fractionally different coordinates, and even the curves are likely to be subtly different.
On the other hand, font names are trademarked, so you can't just use the "usual" name of a font to sell a clone of it.
When I went to maker faires, the only projects I ever saw using Intel's boards were done by students, who obviously got them free or at a very discounted price. Seriously, it seems like Intel's only purpose was "microcontrollers, but USING EX-EIGHTY-SIIIIIIIIIX!"
Their other boards (Edison, etc.) had this tiny-ass connector that was very maker UN-friendly, and not even designed for many attach-remove cycles. And then they refused to release detailed NDA-free chipset documentation; it was supposedly hard even for people who knew actual Intel engineers to figure shit out.
They couldn't make up their mind whether they were going for professional embedded manufacturers with an engineering department, going for makers, or just trying to make an excuse for X86 In Everything. And they managed to fail at all of these.
If it is, then they are making counterfeit chips with ST markings. I bought a bag of 20 bluepill boards a few months ago, and they were all ST-labeled. The other thing is that bluepill boards use the 64K flash version. They all have 128K flash (which is great if you are using mbed, since it matches a supported ST Nucleo board), but the extra 64K isn't certified. It is still very likely that the other 64K works, but it's cheaper for ST to make one die and only certify the other half for customers that want to pay an extra ten cents.
I think the best thing you can do is don't plug anything into the integrated Ethernet port (or on-chip WiFi if it has that). You know that its built-in spyware knows all about how to use that. It's trickier to do that on a laptop, but a USB Ethernet will at least be much more trouble for it to spy on.
The real question is what is Atari bringing new to the console market? The XBox and Play Station market, Are competing for the High End gaming market. Nintendo is mostly focused on the family market.
Think of it in terms of price ranges. I could see it if they are going for the $49-$99 market. Nintendo has the $149-$249 market covered, while Sony and Microsoft have the $250-$399 market covered. But the important question is how do they plan to sell the blades for this razor? The money isn't in selling consoles, no matter how cheap, unless you're Chinese and selling glop-chip consoles loaded with unlicensed games.
SD card slot yes, but does it have DRM like most modern consoles to prevent running unauthorized (and presumably unlicensed) games? Remember, boys and girls, the Atari 7800 was the first system to use a crypto DRM lockout! We can only create new games for unmodified consoles because the master key was found (more than once!) on hard drives dumpster-dived from Atari.
As for the looks, I rather like it. The slotted pyramid part makes it look a bit Daft Punk-ish, like they cut the head off of of an old 2600 and turned it into a robot or something. It feels almost as if 1980's Steve Jobs was involved in the design.
I hate it when people decide that their idea of what UI widgets should look like is so important that they ignore native UI widgets. Doubly so in an app that isn't cross-platform. Cross-platform apps are a minor excuse for avoiding native widgets in favor of the UI framework's widgets, but it's still lazy as fuck.
This is why I use Seamonkey. It is basically what became of the original Mozilla browser after Firefox forked from it. It quickly achieved Firefox 3.x levels of usability and stayed there. Meanwhile, Firefox started chasing Chrome's UI weirdness. I had a chance to install FF yesterday and it was like going to another planet.
The "if it isn't encrypted, we must scare the user" thing is massively stupid. First of all, not everybody runs a web site that needs certificate-based encryption. Yeah, it was good to shame the major players into it, but encryption is hardly important for running a little blog web site on your own domain.
For a small web site, even turning on SSL can be a major undertaking, and getting a certificate is extra work too. I have mine set up aliased to multiple domains, using vhost to select which site to serve up, and AIUI, I would have to get a separate certificate for each domain, and return the proper one for a browser to not bitch.
Certificates (authenticating the server) and encryption (encrypting the communication) are two different things, yet they get conflated by the "we know what's best for you" types.
Oh hell, those damn Windows keys. Both the Windows and "menu" keys steal focus when brushed by accident while touch-typing. It is especially annoying when you're playing a game in full-screen mode and it switches the screen back to the desktop--from just bumping a bottom row key by accident. And then there's the Insert key. What the fuck is the point of overwrite mode in a GUI text editor? Who needs that anymore?
In my previous workplace, I removed various key caps of annoying keys (including Windows, Menu, F1, Insert, and another F-key that did an annoying thing in the IDE) and put them in a small baggie under the monitor. Really, that's the best solution.
Auto update that closes open applications with unsaved work in progress, How hard would it be to send save commands before closing a file and appending a WIP designator to the file name so nothing is over written.
I love how OS X 10.9 is so good about reloading apps after a reboot or kernel panic. I got used to that in my web browser, but it's like magic when (most) apps simply go back to where they were. I guess that's one of the side-benefits of having the high-level UI support built into the operating system. When it gets updated, even old apps can take advantage of new features.
Had to swap around laptops the day before yesterday, moved the 16GB over. Had three problems in 24 hours (two failures to wake from sleep and one kernel panic), so popped it open, took out the RAM, pencil erasered the SODIMM contacts and brushed them off, put them back in tight. I've been doing this since SIMMs back in the '90s, I remember one Mac SE that would spontaneously reboot until I cracked it open and re-seated its memory.
I don't know much of any good reason to use 10.7 or 10.8.
10.4 on a PPC system lets you run Classic. (Except that my recent Uverse modem does some shit that fucks up Netinfo on 10.4, which sucks.)
10.5 lets you work with bigger disks, and is the last PPC version.
10.6 lets you run Rosetta so you can still run PPC apps on Intel. I ran this for a long time on my Late 2011 MacBook Pro, which required a tricky process to downgrade since the shrinkwrap version only went up to 10.6.3., and it couldn't run anything earlier than 10.6.7. Had to get a hacked up 10.6.7 image from TPB, then run the 10.6.8 updater.
10.7... um... it removed Rosetta?
10.8 is the last one to support Open Transport (bye-bye MT Newswatcher in 10.9)
10.9 fixed up problems with OpenGL that would let programs (specifically Minecraft 1.6) totally hose the graphics subsystem. That's what got me to upgrade to 10.9.5 two years ago.
10.10 made the Notification Manager more complete, yeah, whatever.
10.11 locked down the system a lot, but at least you can still do magic to turn that off
10.12 locked things down even more
So I run 10.9.5 on my "daily driver", leave other Mac Minis at whatever was on when I got them (10.9 - 10.11), and I put 10.6.8 on an older MacBook Pro (one of the first Core 2 Duo models) when I need some old stuff to work.
Unfortunately they also didn't pop it up as the top window on your desktop.
That brings up a big complaint of mine about WIndows. My brother's old laptop had a few problems (bad connection to DC plug, and the hard drive was all clickety-clickety dead), so I had to re-install W7 ("Home Premium", gag, but that's what the CoA sticker said, so that's what I downloaded from TPB) on a spare spinny HD I had lying around. I had forgotten how bad Windows was about putting up new alert windows behind everything else, usually completely hidden by the Explorer window that you started something from. The stupid security dialogs from running the various driver installers almost always went to the back, and even then the app's window would sometimes go in back. Thank goodness for the task bar.
At least I didn't get the really annoying "modal child window alert behind parent window" thing. Oh sorry, you can't move this window on top to get to the window it's hiding, because it has a child window, the one that it's hiding!
Oh yeah, and I think Windows still doesn't fucking erase the background of exposed windows when one is closed (something Mac has done since 1984), so it looks like windows are falling through each other before the behind windows paint themselves. At least It's not so bad on modern graphics cards where each window gets its own buffer and the GPU handles putting them on the screen.
I have a USB-based EPROM programmer (Needhams EMP-31?) which only has a 32-bit USB driver. WTF? The company was bought out in early 200x just before x64 became a thing, and the big fish company simply threw their whole product line out the back door. This was also before libusb and "user-land drivers for everything on USB" became a thing. (Even then, Windows might still want INF-only "driver" to tell it to fuck off and leave the thing alone instead of installing its own retarded driver. AIUI, Windows has had a crappy CDC serial driver at least through Windows 7.)
From what I have heard, the base reason for this WTF was that it probably emulated a bit-banged parallel port over USB in a way that required a Windows driver. So because of how it hooked itself into Windows, it had a driver incompatible with 64-bit mode. I now keep it together with a crappy old white-era Dell Insprion with 32-bit XP installed.
Windows started with a brain-damaged USB subsystem that wants to freaking re-install a driver completely if you merely plug a device into another port. Sometimes that's useful when a particular driver instance gets fucked. Windows has maintained that brain damage for over 15 years since then.
So yeah, my gripe is that Windows has always had crap USB support. At least I normally use OS X, so I only have to use Windows (which I still keep at Windows 7) for playing games.
I work somewhere where 3D TV was part of our graphics card driver validation process, and that was even worse: It was like cardboard cutouts being moved on top of or behind each other.
That's because of the "cheap way" to do 3D: layer a bunch of 2D images in 3D space, tilted at slightly different angles. Full dual-camera filming and full 3D computer rendering (and both when both are used) should not have this problem. But really 3D is an answer in search of a question. It's amusing every now and then as a novelty, but nobody cares enough to go out of their way for it. And there are a not insignificant percentage of people for whom the effect doesn't even work properly.
I don't even need super resolution, 720p is plenty enough unless you want to see how long it's been since an actor shaved that morning, and I usually don't have a problem watching low-bit-rate 480p on an ATSC sub-channel.
Yeah, okay, you're ranting off-topic because you got triggered by the N-word, but I have to laugh. Years ago I bought a RadioThermostat model resold by 3M (for $100), and even lucked into a second one later for almost nothing. First thing I did was hook up the stupid C wire; both houses I installed them in were made in the '60s and had a five-wire cable to the thermostat, with the blue wire not connected. The other thing that Nest apparently failed to do was make it easy to tell when the thermostat had stopped talking to the server. The web page for my thermostat shows the time of last contact, which happens every 5 minutes or so.
It also uses a cheap segmented LCD and doesn't run fucking Linux, just a couple of low-power micro-controllers. If it weren't for the WiFi radio, it could run a year on three AA batteries. It's just a fucking thermostat, not some kind of "home comfort hub" bullshit. I don't need it to "learn" anything, I just need a simple schedule and an away setting. Oh yeah, and it's even possible to control it over your LAN with JSON requests and to change the "cloud" server URL if you don't want to use theirs.
Too bad I live in the middle of Texas, where it doesn't get cold enough for that "rig a fail-safe mechanical thermostat in parallel" tip to be useful.
The standard HID configuration for keyboards (the one that works with BIOS at boot time) has one byte of 8 modifier keys, plus 6 bytes for other keys. This limit comes from the "slow" 1.5Mbit speed of USB only allowing 8 bytes of data in a transaction. (I can't remember right now what the other byte is.) There are no key up or key down events, just a list of currently pressed keys. Anything beyond that requires the host to use a new HID configuration with a larger data response, and presumably needs the 12Mbit speed as well.
The idea in the previous reply about pretending to be multiple keyboards (probably as a composite device) sounds like an interesting new approach to the problem.
If true, the reason is likely a lack of diodes in the matrix to prevent ghosting. Diodes are required for true N-key rollover, but it should be possible to have pseudo-N-key rollover by allowing any number of keys from the same row or column, and stop generating keypresses when a ghost happens. (Modifier keys are usually on their own inputs to keep them out of the matrix.)
In my experience, an "authentic" Model M has an RJ-style connector with the clips on the side instead of the top. Back in the day I got quite a few of both AT and PS/2 cables for them. The AT connector just needs a simple dongle adapter to PS/2. And of course they have no Windows keys.
But I'd rather just replace its controller board with a micontroller to make a USB version.
Slashdot Mirror
I had the older version of their OTA DVR. It was sad when Rovi (the company that brought us Macrovision) suddenly shut down the OTA TV Guide service. It was great having two weeks of guide data. Now you're lucky to get 24 hours. Why? Because TV set manufacturers suck, and some older (and probably not-so-old) TV sets can't handle that much guide data, so stations don't send much.
I just checked, and all but three of my local stations have 24 hour guide data. Of the other three (at least 8 sub-channels total) one has 36 hours, and the other two have 72 hours of data.
Sure, I could subscribe to a feed for my MythTV, and reading the EPG data on my Hauppage cards sometimes hoses it such that I have a cron job to detect it and reboot, but free is free, dammit.
Some USB receivers have mini antennae, but I've never found them to work in any location, not even in the countryside or a downtown hotel.
I recently read that the reason they don't work well is that they need a ground plane. Notice how they have a really nice magnet? You need to stick that to a big piece of metal, like the roof of a car, for the antenna to work properly. Otherwise you might as well just jam a long piece of wire (like 2m-5m) into the connector and move it around until you get good reception.
Under US copyright law, the "strike" (image) of a font can not be copyrighted. Computer outline fonts can be copyrighted because the outlines are considered a computer program, even more so when there is hinting. Note that this copyright is on the specific outlines of a font. If you make a clone that looks the same, it will likely at least have fractionally different coordinates, and even the curves are likely to be subtly different.
On the other hand, font names are trademarked, so you can't just use the "usual" name of a font to sell a clone of it.
When I went to maker faires, the only projects I ever saw using Intel's boards were done by students, who obviously got them free or at a very discounted price. Seriously, it seems like Intel's only purpose was "microcontrollers, but USING EX-EIGHTY-SIIIIIIIIIX!"
Their other boards (Edison, etc.) had this tiny-ass connector that was very maker UN-friendly, and not even designed for many attach-remove cycles. And then they refused to release detailed NDA-free chipset documentation; it was supposedly hard even for people who knew actual Intel engineers to figure shit out.
They couldn't make up their mind whether they were going for professional embedded manufacturers with an engineering department, going for makers, or just trying to make an excuse for X86 In Everything. And they managed to fail at all of these.
If it is, then they are making counterfeit chips with ST markings. I bought a bag of 20 bluepill boards a few months ago, and they were all ST-labeled. The other thing is that bluepill boards use the 64K flash version. They all have 128K flash (which is great if you are using mbed, since it matches a supported ST Nucleo board), but the extra 64K isn't certified. It is still very likely that the other 64K works, but it's cheaper for ST to make one die and only certify the other half for customers that want to pay an extra ten cents.
I'm sure we'll be seeing them sponsor a law to shorten the length of patents any day now.
I think the best thing you can do is don't plug anything into the integrated Ethernet port (or on-chip WiFi if it has that). You know that its built-in spyware knows all about how to use that. It's trickier to do that on a laptop, but a USB Ethernet will at least be much more trouble for it to spy on.
The real question is what is Atari bringing new to the console market? The XBox and Play Station market, Are competing for the High End gaming market. Nintendo is mostly focused on the family market.
Think of it in terms of price ranges. I could see it if they are going for the $49-$99 market. Nintendo has the $149-$249 market covered, while Sony and Microsoft have the $250-$399 market covered. But the important question is how do they plan to sell the blades for this razor? The money isn't in selling consoles, no matter how cheap, unless you're Chinese and selling glop-chip consoles loaded with unlicensed games.
SD card slot yes, but does it have DRM like most modern consoles to prevent running unauthorized (and presumably unlicensed) games? Remember, boys and girls, the Atari 7800 was the first system to use a crypto DRM lockout! We can only create new games for unmodified consoles because the master key was found (more than once!) on hard drives dumpster-dived from Atari.
As for the looks, I rather like it. The slotted pyramid part makes it look a bit Daft Punk-ish, like they cut the head off of of an old 2600 and turned it into a robot or something. It feels almost as if 1980's Steve Jobs was involved in the design.
How about this: rewrite systemd in Rust first and I might consider it.
The battery menu on at least 10.9.5 will try to identify power-hog apps and list them for you.
I hate it when people decide that their idea of what UI widgets should look like is so important that they ignore native UI widgets. Doubly so in an app that isn't cross-platform. Cross-platform apps are a minor excuse for avoiding native widgets in favor of the UI framework's widgets, but it's still lazy as fuck.
This is why I use Seamonkey. It is basically what became of the original Mozilla browser after Firefox forked from it. It quickly achieved Firefox 3.x levels of usability and stayed there. Meanwhile, Firefox started chasing Chrome's UI weirdness. I had a chance to install FF yesterday and it was like going to another planet.
The "if it isn't encrypted, we must scare the user" thing is massively stupid. First of all, not everybody runs a web site that needs certificate-based encryption. Yeah, it was good to shame the major players into it, but encryption is hardly important for running a little blog web site on your own domain.
For a small web site, even turning on SSL can be a major undertaking, and getting a certificate is extra work too. I have mine set up aliased to multiple domains, using vhost to select which site to serve up, and AIUI, I would have to get a separate certificate for each domain, and return the proper one for a browser to not bitch.
Certificates (authenticating the server) and encryption (encrypting the communication) are two different things, yet they get conflated by the "we know what's best for you" types.
Oh hell, those damn Windows keys. Both the Windows and "menu" keys steal focus when brushed by accident while touch-typing. It is especially annoying when you're playing a game in full-screen mode and it switches the screen back to the desktop--from just bumping a bottom row key by accident. And then there's the Insert key. What the fuck is the point of overwrite mode in a GUI text editor? Who needs that anymore?
In my previous workplace, I removed various key caps of annoying keys (including Windows, Menu, F1, Insert, and another F-key that did an annoying thing in the IDE) and put them in a small baggie under the monitor. Really, that's the best solution.
Auto update that closes open applications with unsaved work in progress, How hard would it be to send save commands before closing a file and appending a WIP designator to the file name so nothing is over written.
I love how OS X 10.9 is so good about reloading apps after a reboot or kernel panic. I got used to that in my web browser, but it's like magic when (most) apps simply go back to where they were. I guess that's one of the side-benefits of having the high-level UI support built into the operating system. When it gets updated, even old apps can take advantage of new features.
Had to swap around laptops the day before yesterday, moved the 16GB over. Had three problems in 24 hours (two failures to wake from sleep and one kernel panic), so popped it open, took out the RAM, pencil erasered the SODIMM contacts and brushed them off, put them back in tight. I've been doing this since SIMMs back in the '90s, I remember one Mac SE that would spontaneously reboot until I cracked it open and re-seated its memory.
I don't know much of any good reason to use 10.7 or 10.8.
10.4 on a PPC system lets you run Classic. (Except that my recent Uverse modem does some shit that fucks up Netinfo on 10.4, which sucks.)
10.5 lets you work with bigger disks, and is the last PPC version.
10.6 lets you run Rosetta so you can still run PPC apps on Intel. I ran this for a long time on my Late 2011 MacBook Pro, which required a tricky process to downgrade since the shrinkwrap version only went up to 10.6.3., and it couldn't run anything earlier than 10.6.7. Had to get a hacked up 10.6.7 image from TPB, then run the 10.6.8 updater.
10.7... um... it removed Rosetta?
10.8 is the last one to support Open Transport (bye-bye MT Newswatcher in 10.9)
10.9 fixed up problems with OpenGL that would let programs (specifically Minecraft 1.6) totally hose the graphics subsystem. That's what got me to upgrade to 10.9.5 two years ago.
10.10 made the Notification Manager more complete, yeah, whatever.
10.11 locked down the system a lot, but at least you can still do magic to turn that off
10.12 locked things down even more
So I run 10.9.5 on my "daily driver", leave other Mac Minis at whatever was on when I got them (10.9 - 10.11), and I put 10.6.8 on an older MacBook Pro (one of the first Core 2 Duo models) when I need some old stuff to work.
Unfortunately they also didn't pop it up as the top window on your desktop.
That brings up a big complaint of mine about WIndows. My brother's old laptop had a few problems (bad connection to DC plug, and the hard drive was all clickety-clickety dead), so I had to re-install W7 ("Home Premium", gag, but that's what the CoA sticker said, so that's what I downloaded from TPB) on a spare spinny HD I had lying around. I had forgotten how bad Windows was about putting up new alert windows behind everything else, usually completely hidden by the Explorer window that you started something from. The stupid security dialogs from running the various driver installers almost always went to the back, and even then the app's window would sometimes go in back. Thank goodness for the task bar.
At least I didn't get the really annoying "modal child window alert behind parent window" thing. Oh sorry, you can't move this window on top to get to the window it's hiding, because it has a child window, the one that it's hiding!
Oh yeah, and I think Windows still doesn't fucking erase the background of exposed windows when one is closed (something Mac has done since 1984), so it looks like windows are falling through each other before the behind windows paint themselves. At least It's not so bad on modern graphics cards where each window gets its own buffer and the GPU handles putting them on the screen.
I have a USB-based EPROM programmer (Needhams EMP-31?) which only has a 32-bit USB driver. WTF? The company was bought out in early 200x just before x64 became a thing, and the big fish company simply threw their whole product line out the back door. This was also before libusb and "user-land drivers for everything on USB" became a thing. (Even then, Windows might still want INF-only "driver" to tell it to fuck off and leave the thing alone instead of installing its own retarded driver. AIUI, Windows has had a crappy CDC serial driver at least through Windows 7.)
From what I have heard, the base reason for this WTF was that it probably emulated a bit-banged parallel port over USB in a way that required a Windows driver. So because of how it hooked itself into Windows, it had a driver incompatible with 64-bit mode. I now keep it together with a crappy old white-era Dell Insprion with 32-bit XP installed.
Windows started with a brain-damaged USB subsystem that wants to freaking re-install a driver completely if you merely plug a device into another port. Sometimes that's useful when a particular driver instance gets fucked. Windows has maintained that brain damage for over 15 years since then.
So yeah, my gripe is that Windows has always had crap USB support. At least I normally use OS X, so I only have to use Windows (which I still keep at Windows 7) for playing games.
I liked it better when it was The Music Man IN SPAAAAAACE!
Few actors get such a good last role.
I work somewhere where 3D TV was part of our graphics card driver validation process, and that was even worse: It was like cardboard cutouts being moved on top of or behind each other.
That's because of the "cheap way" to do 3D: layer a bunch of 2D images in 3D space, tilted at slightly different angles. Full dual-camera filming and full 3D computer rendering (and both when both are used) should not have this problem. But really 3D is an answer in search of a question. It's amusing every now and then as a novelty, but nobody cares enough to go out of their way for it. And there are a not insignificant percentage of people for whom the effect doesn't even work properly.
I don't even need super resolution, 720p is plenty enough unless you want to see how long it's been since an actor shaved that morning, and I usually don't have a problem watching low-bit-rate 480p on an ATSC sub-channel.
And whatever you do, be sure not to play this in your house!
Yeah, okay, you're ranting off-topic because you got triggered by the N-word, but I have to laugh. Years ago I bought a RadioThermostat model resold by 3M (for $100), and even lucked into a second one later for almost nothing. First thing I did was hook up the stupid C wire; both houses I installed them in were made in the '60s and had a five-wire cable to the thermostat, with the blue wire not connected. The other thing that Nest apparently failed to do was make it easy to tell when the thermostat had stopped talking to the server. The web page for my thermostat shows the time of last contact, which happens every 5 minutes or so.
It also uses a cheap segmented LCD and doesn't run fucking Linux, just a couple of low-power micro-controllers. If it weren't for the WiFi radio, it could run a year on three AA batteries. It's just a fucking thermostat, not some kind of "home comfort hub" bullshit. I don't need it to "learn" anything, I just need a simple schedule and an away setting. Oh yeah, and it's even possible to control it over your LAN with JSON requests and to change the "cloud" server URL if you don't want to use theirs.
Too bad I live in the middle of Texas, where it doesn't get cold enough for that "rig a fail-safe mechanical thermostat in parallel" tip to be useful.
The standard HID configuration for keyboards (the one that works with BIOS at boot time) has one byte of 8 modifier keys, plus 6 bytes for other keys. This limit comes from the "slow" 1.5Mbit speed of USB only allowing 8 bytes of data in a transaction. (I can't remember right now what the other byte is.) There are no key up or key down events, just a list of currently pressed keys. Anything beyond that requires the host to use a new HID configuration with a larger data response, and presumably needs the 12Mbit speed as well.
The idea in the previous reply about pretending to be multiple keyboards (probably as a composite device) sounds like an interesting new approach to the problem.
If true, the reason is likely a lack of diodes in the matrix to prevent ghosting. Diodes are required for true N-key rollover, but it should be possible to have pseudo-N-key rollover by allowing any number of keys from the same row or column, and stop generating keypresses when a ghost happens. (Modifier keys are usually on their own inputs to keep them out of the matrix.)
In my experience, an "authentic" Model M has an RJ-style connector with the clips on the side instead of the top. Back in the day I got quite a few of both AT and PS/2 cables for them. The AT connector just needs a simple dongle adapter to PS/2. And of course they have no Windows keys.
But I'd rather just replace its controller board with a micontroller to make a USB version.