Posts by Henry Wertz 1 3450 publicly visible posts • joined 12 Jun 2009
Yeah of course there'll be ongoing exponentoial growhtj forever *rolls eyes*. The cell cos got into a tizzy years back over this, before realizing people could only watch so many cat videos in a day, usage plateaued (to the point that they now sell 5G Home Internet widely to use that excess capacity the phones aren''t.).
Flock only claims to own the info from the cameras to evade privacy laws. If a state, county, municipality says the police or government must not run surveillance, Flock claims the cameras are run on behalf of the police or local gov't.. If laws restrict private companies from doing so, they claim they provide the service but the local muiniciplaity is running them and it's their info.
Yeah in a lot of cases it's like an agrressive, power hungry lint. Here's a potential flaw! Is it even potentially possible to exploit? Often the answer is 'no'. I saw a flaw brought up last week, clear buffer overflow, fixed size variable but length check whatsoever at that point in the code. But unexploitable in practice because a bounds check *was* done by the caller of that function, and the shoddy function was not directly callable otherwise. I'm sure this code has been fixed but it's one of those ~98% with no exploit.
'Parra claims the new MCP will be easier to deploy and use'...
Then just later discuss how it's harder to set up, and will be a slog. I mean, use for the END user is pointing your MCP client at an MCP server.
I just thought that was amusing. The actual changes sound solid. i suspect MCP design was VERY quick and dirty (like they just stuck json serialize/deserialize straight into the code as opposed to desigining a protocol as such) so this is probably well needed.
Don't know if you're serious or not (I really despise the out of the box UI on current Ubuntu versions...) but Windows' shell can be replaced, there's things like SeelenUI with plus theme and icon pack that can apparently look pretty close. I swirched to Linux in 1993 though, so I only tried an alternate shell in the XP era, on a install I booted into.a few timea a month at first, then didn't boot into at eventually.
But AFAIK you can still replace the shell and (surprisingly given how spaghetti-code windows seems to be, i.e, how tied together components are) apparently the shell maintains it's seperation and switching it out is a smooth and trouble-free process.
I use KDE myself, but my dad uses gnome flashback. He was used to Windows, then to the ~Ubuntu 12 (2012-era) desktop. Flashback looks just the same. He COULD learn a new desktop, but there's literally no reason to. There's something to be said about being able to use the latest and greatest apps without having your UI change on you every couple years.
Microsodt apparently forgot they used to have autorun in CDn and USB device insertion... but turned it off due to it's possibility for being abused. Although I doubt anyone expected a MONITOR of all things to abuse it.
I must add this makes me glad I swirched to Linux decades ago (1993). The only ads I've gotten from any of my distros, ever, are the mild things telling you how great the apps are on screens when the installer was running, and a ONE TIME notice that I could do backups with 'Ubuntu One' (which has been replaced with a general one time notice suggesting some kind of backups be set up since they discontinued their backup service.). And if even that is too much for one's sensibilities. well, there are even more 'libre' distros where they'd consider even that a step too far.
Americans are dumb (disclaimer: I live in USA so I say this from first hand experience). so adding additional milk proteins and whatever making fortified milk, people were like 'Oh hell no, we won't buy milk with stuff addeed to it.'. The 'ultrafiltering' takes like 1.5 units of milk and filter out the 'extra' water so you end up with the same end result. But the same people who woiuldn't buy the first product gleefully buy the second even though the result is the same.
As for whether one NEEDS this fortified milk, that's another issue.
This. Since I've been using Linux since 1993, I can say for running Linux of that era, you had UNIX workstations with 680x0 chips where they made 68020s with a 68851, Like the Sun 3 and contemporaries from HP etc. The companies that made their own chips like HP PA-RISC, SGI MIPS, Sun SPARC, etc., all came out in the late 1980s so until then a lot of them used 68K chips. So you had Sun3 systems etc. that had 68020+68851 and they were supported. (The 2 biggest UNIX workstation vendors that didin't run a 68K were DEC and IBM, they both had 1970s-era designs to hold them while they got their modern 64-bit chips ready (IBM POWER/PowerPC and DEC Alpha), when the DEC Vax finally ran out of steam the Alpha wasn't quite ready, they used MIPS instead of a 68K in that interim.)
68020, the Amiga, Atari ST did not come with a 68851 socket, and the Mac II was the only Mac that had a socket; but it didn't come with a 68851 in it unless you ordered one with A/UX (a UNIX Apple sold for a few years back then). So in reality for 1980s/1990s desktops, you make sure to get one with a 68030 although the 040 was apparently a fair bit faster. I have no idea if any of the systems back then shipped with a 68060, if so it was not common.
And, uclilnux. It's been around for many years, a MMUless branch of the Linux kernel. 4 or 5 years ago, this was merged into the mainline Linux kernel, you set "MMU=no" in the kernel config and you've got a uclinux kernel running. That'll run on anyhting 32-bit basically. ELKS runs down to like 8088s etc.,. they stripped it down so you can run some semblance of Linux on a 256KB system.
Saw a drivethrough sign here pack up a few years ago, and it was running Ubuntu 12.04. Amusingly with their extended support they could be running 12.04 still and STILL have security updates for it (after the 5 years of regular support one can pay to extend by 10 years so it'd still have almost a year of updates.)
Could they have run reduced service at least with ETCS, LZB, and manual signalling? Most likely. But keep in mind this was only a 55 minute outage. They may have found that actually slowed things down (since GSM-R coming back up, they'd have these manually dispatched trains in mid-transit, and not in the planned locations. At worst case they'd have to wait for them to finish, at best case they would probably still have some in 'the wroing location' to just start them up on the regular schedule.
I actually had the opposite problem back in the day. I started using Linux in 1993... I was running 800x600 in 256 colors, or 1024x768 1-bit (the card was actually running 4-bit -- 16 collor -- mode, but this was unheard of in UNIX so if you actually went to run any software in 16 color mode it almost always immeidately crashed. So XFree86 allowed presenting 16 color modes as monochrome.)
But by the late 1980s, it was completely routine for these UNIX workstations to have 21" CRT running at 1280x1024 or even 1600x1200. They figured if you were already spaffing out 5-10K for a system, of course you'd spend like $500-1000 for a large high-resolution monitor. My first year or two of using Linux, when I did fire up X to run some application, probably half the time it woudn't even fit on the screen. So ironically the early work on Linux GUI applications amounted to trimming them down so they could run on the low-res screens typical of the PC.
As for Access... for those who really haven't used it... it's like a form builder on top of a sqlite-style thing (in terms of the Access 'database' just being a single file that it read and wrote to directly) but Microsoft Jet had none of the speed, stability, robustness, error recovery, or crash recovery of sqlite. They did apparently replace this with "ACE" in 2007 (if you have an .mdb file it's JET, if you have .accdb it's ACE). Which I assume is better in some way than JET? I don't know. Basically it was quick and easy to rig a form and get something up and running; it was really not very good to be honest but (as you see in the comments) the same places that abuse Excel spreadsheets for things would almost invariably have been using Access at some point if not still.
That 22" form limit limitation is really weird. Given (in the UNIX world at least) 21" monitors were already common, having a limit of 22" doesn't seem that forward thinking. I'm also just not sure why there'd be a size limit at all. I bet it's something weird like they used some arbitrary units and that turned out to be 32768 units (limit of a signed 16-bit integer) or some odd thing like that. I'll just say... a) Better late than never I guess? b) I'd hate to see what someone is coming up with that really needs ore than a 22" wide form. That's going to be one heinous Access application LOL.
Have a Surface Pro 5, that my sister managed to trash Windows on so badly that the screen was stuck upside down, and it spent like 5 minutes on every startup and shutdown trying to install some Windows updates, failing, then rolilng back. Runs great with Ubuntu, and I threw on the surface kernel so the touch screen and camera work. I guess I'll make sure not to probe the SAM then LOL.
I'll note, I don't have problems adjusting the brightness on mine with the brightness keys, or the sliders elsewhere in the GUI; so presumably the Surface kernel already knows how to do the brightness, Copilot could have looked through there for how to do it presumably. Ahh well, I guess blindly probing the hardware works except when it doesn't LOL.
my friend had a 1988 Buick (in the 2000s) that was acting a bit wierd. Indeed, the ECU for 1 or 2 model years was prone to cracking, which would needless to say cause ALL KINDS of fun issues. It ran surporisingly well given it was randomly losing signals, instead of dying it'd at least go into some limp home mode so it just ran a bit crap. He just ditched it for another vehicle.
no I think youlre spot on,. And I don't see it as semantics. The first solution is probably descended from chroot jails (with some functionality to let it be better isolated, have 'fake root', cgroup type stuff for resource control, etc.). The other is just a VM. I think the only thing 'micro' about it is just that a system that doesn't load mich software doesn't use much RAM.
Apple can call this a lightweight container the way Apple of the 1980s called the cooperative task switching of macOS 6-9 'multitasking', but it's really not. The result works but it's really not as elegant.
don't know why you got thumbs downed. The article says it uses virtualization framework, and the bit about them at present allocating memory on use but not returning it to the system on free should give a hint to readers that it's not what literally everyone else calls a lightweight container, it is in fact a VM.
Google'ing it confirms this.
Yes that's a description of FeeeBSD containers. But the article mentions memory being grabbed on use but not released, and the use of Apple's virtualization framework. In other words I don't think this is using what everyone else calls lightweight containers, it's using VMs.
I'll just say, physically showing up and copying stuff onto USB sticks is definitely bold. If they get caught hacking in, they could just as well be on the other side of the planet. Here, if they get caught, the doors can be locked and the cops come by and physically pick them up, they are rumbled.
To be honest, carriers brought on the situation with 5G themselves. They had data only working, with voice coming out in 6 months to a year.., and carriers were like 'You know? We'll just start putting in 5G now.'. Yeah, if you start throwing in gear before the standard is all standardized you end up with more pain than if you wait just a bit.
The article says signals drop off quickly outside urban areas, but that's simply not the case. If you travelled here from Europe like 10 or 15 years ago and saw this, GSM coverage was poor for a number of years, because Verizon (the coverage king for many years) ran a CDMA network (as did Sprint.. the less said about them the better), in my area US Cellular also used CDMA. AT&T did a rather poor GSM rollout at first (1900mhz band only, on sites spaced for 800mhz) and T-Mobile back then did not focus on coverage outside urban areas, and also ran a 1900mhz-only network.
Well, T-Mobile bought 600 and 700mhz spectrum; AT&T rolled out on 800mhz as well as 700mhz (among other bands for both); Verizon still has great coverage; and all the carriers here have fully upgraded to 4G LTE (and a lot of 5G now too) so the "two incompatible network standards" is a thing of the past. I've driven from where I live ~800 miles south to New Orleans (south coast), to the east coast (~1000 miles) and west coast (~1000 miles) and have continuous coverage. Not just on the interstates, I've gone through the back roads... still continuous coverage.
You can look at coverage maps and see these big uncovered areas in the western US, but it's literally high desert with no roads, walking trails, or anything going through them.
Just saying, this technology is cool. But, personally even if I travelled around the country full time, I would not pay a penny for satellite backup service. In the last 10 years, I've probably only been out of cellular coverage for an hour or two total. The main area I've had bad coverage is in SW Wisconsin, the roads go through finger valleys; but, in those spots you tend to have rock faces on several sides of you so your view of the sky is also pretty obstructed (i.e. I'm not at all sure that satellite backup service would work either.) Even up there, the dead zones seem to be smaller each time I visit as the cell cos put up 600 and 700mhz coverage (which has a bit better range) and I think just plain cranked the power up a bit on the cell sites in these areas to 'brute force' some coverage into those hard-to-reach spots.
They MIGHT get some interest in Alaska, for instance. But a lot of the US is very well covered.
I'm surprised they don't port Nutanix to ARM. They have a KVM-based hypervisor, a distributed block storage system, and a lot of 'glue' to run a Kubernetes-style setup, and easily move between their setup and various other cloud setups. I suppose if they have things extensively customized a port to ARM could be tricky, but I just would not be surprised they'd find out it's not that difficult if they pursued it.
Hopefully ones do come out with Linux. I was waiting to get a Qualcomm one, but they've inexplicably decided they are not interested in having Linux run well on the Snapdragon X etc. Nvidia of course has been running Linux on their ARM setups for many years. Zero interest in Windows11. I ran Acer CB5-311 Chromebook with a Tegra K1 under 'Chrubuntu' (installed Ubuntu but kept the kernel and nvidia drivers from ChromeOS) and it was FANTASTIC. Fast and excellent battery life (22 hours under typical use and 12 hours doing video encodes so all cores were maxed out.). During video playback it could (and in fact did) actually fully power off the CPU *between* video frames since wakeup from fully off was under 1/1000s of a second.
If it's good enough for Intel... I'll just comment Intel referrs to their 'enhanced superfin' 10nm process as 'Intel 7' (implying it's equivalent to a 7nm process) and renamed their 7nm process 'Intel 4.'.. With the aame reasoning (we can do all this adnvanced stuff to get more speed and desnity out of the process). Of course that Intel 4 being comparable to 4nm is assuming those actually producing 4nm chips wouldn't just use similar methods themselves.
So although I thinkl it's silly for Huawei to do this, Intel pulls similar shenanigans with marketing their foundry tech.
Amusingly CenturyLink (direct descendent of USWest, another 'regional bell operating company' from the Bell breakup that covers the has service in the central US), when my parents used to have a landline it'd flake out more than the cell phones. Not a line fault, fast busies and 'all circuits are busy' and calls just not going through from time to time. They had a power outage downtown once that took their DSL down for a day or so -- I ASSUME the voice service was battery backed (following Bell standards), maybe, but no battery backups in the central office for the DSL.
Verizon Wireless, in contrast, has battery backup on all sites as far as I know; generators on their "bigger" sites; and point to point microwave link fallback so if someone cuts the fiber or whatever that's running to the site the service stays up (albeit possibly slower, I doubt the microwave links have the same capacity as the fiber line. But that depends on how much traffic is on the site of course.) We had a tornado come through, my house kept power but everything from about a block away to downtown lost power; the cell phone service blipped off for about 10 seconds and dropped my call (I guess the batteries must have been dead and the generator didn't kick on fast enough? LOL) then service was normal after that. Well, it was terrible downtown since several of the cell sites got destroyed, outside downtown even in the "no power" areas it was totally normal. The landline service (central office is downtown) was dead for days.
Problem is. the fiber will not be rolled out in all the areas they retire the copper.. And (assuming your househiold is not in a dead spot), they don't have comparable plans and prlcing on the wireless side. They've done this in other states, just withdrawn trhe copper but not rolled out fiber to replace it. And their 5G pricing is a joke so moving the DSL customers to 5g doesn't help wioth affordability.
I'd say (given they really do have few customers) they'd have zero complaints if they juist took a technology-agnostic approach, provide one of those cellular to POTS boxes for those who are using landline, and 5G box rate limited to the speed they paid for on DSL but same price for people with DSL,
I'm surprised at the number of thumbs down. Having a multi-master (i.e. there's not some definitive central database that changes come out from) directory service propogate changes is not the simple problem it'd appear to be at a glance. At the extreme end, 'instant' propogation means a site has to check other sites to make sure a directory entry (API key in this case) is still valid, which defeats the purpose of it being distributed. And in general, the more frequenty these synchronize the higher the overhead is. It's really not surprising there's some propogation delay; 23 minutes seems a bit high but 5 or 10 would not have surprised me one bit.
That said, if Google finds rapid directory sync in the general sense to be infeasible.. in this specific case of deleting API keys, Google REALLY should just suck it up and get them to propogate MUCH faster than they are now. Maybe even create a 'delete API key" API (an API to delete API keys.. heh) that operates outside the way they otherwise synchronize this information specifically so API keys can be deleted rapidy. This may not have been viewed as a big problem in the past, but given how fast these bad actors can rack up a AI slop bill it really is (... I honestly assume when someone takes these API keys and racks up a huge AI bill in a matter of minutes or hours, that they are probably having it churn out AI slop to put on Youtube or Tiktok, in the hopes something gets enough views to get monetized. Generally, crap monotone AI voice, stuff that is mispronounced, repetitive, and showing stock footage that may or may not have anything to do with what they are talking about (i.e. if it was some slop about the country Turkey it may just mix in some photos of thanksgiving turkey dinner, or live turkeys, while it's at it.)
I would have considered just removing the device check. It's clever and a good idea to have one, bur then the executives checking on their Androd tablets will see the Apple Pay choice and be inpressed at how quickly it was implemented.
Don't put the change up like an hour later, this would make one wonder if the work was rushed (since they don't know you're just making the button visible on their device...) but maybe a few days (or a week or two if your company culture or development cycle or whatever makes it so changes take a while to make it to the live site.)
Probably shouldn't be using it anyway. I mean, forget Google suspending their account. Is this control plane running all the time? Yes? Well, then, any cloud provider is intending to make a profit, so they would be able to provide the same service on their own for less than what they are paying Google. Economies of scale are a thing, I can rent a tiny fraction of a server for less than buying one for myself and getting high speed internet to it; but at $10 million spend they are well past that point. And bursty workloads of course, if you only need some capacity now and then, but it sounds like this is probably running all the time so that doesn't apply either.
Beat me to it. I was going to post something very similar. Indeed, if I read 'tracks all AWS services automatically' I expect it to track all AWS services automatically,. Not all but a few, and not some weaselty 'well a marketplace isn't a service.'. They can say 'most' if that's the case, or 'some exceptioins apply' or the like.
It's highly unlikely they will make VMs available.
Apple added some clauses in macOS 11 (Big Sur, so 2020) saying "You may lease or sublease a validly licensed version of the Apple Software in its entirety to an individual or organization"... given... "each lease period must be for a minimum period of twenty-four (24) consecutive hours."
So, the terms prohibit slicing up a large computer into multiple VMs. Some people at the time mulled getting extra macOS licenses, if you have 5 licenses you could run 5 VMs since you have 5 licenses to lease out. But macOS is not sold seperately. And, since the requirement is for a minimum of 24 consecutive hours, you don't get to just fire up a VM for a few minutes or an hour.
I imagine this will NOT be cheap -- the rate for fairly pedestrian Macs in the past was $1 an hour (and again with a 24 hour consecutive minimum, so $24 a throw minimum.) Although I know you know what "consecutive" means, just want to stress again, someone who may even want 24 hours of use throughout the month, but it's an 4 hours in 6 seperate sessions, they will have to pay for 144 hours of usage.
Isn't it illegal to have someone set a spending cap, then automatically raise it unilaterally without telling the customer? If not it certainly should be.
The rest of this is pretty bad too (enabling extra services on what were maps only api keys), but if people were hitting a $250 cap... well it's still bad when someone's expecting a like $50 bill but nothing like getting one for $10,000+.
So, the fire's a fire. Unless IBM's hardware started it there's likely nothing IBM could do about it. But having the status page not reflect a complete shutdown ideally within seconds (but at least within like 5 minutes), that's truly inexcusable and I would ditch a cloud provider that doesn't have a working status page. At that point how much is it a 'cloud' (where your stuff could move from one machine to another and supposedly there's resiliency and failsafes to some extent.) and how much are you just paying for a virtual private server (you have a fraction of a rack mounted server.. usually cheaper than cloud since they don't claim to give you anything more than that.)
As a Linux user since 1993, I had (up to a few years ago) been wondering if they were just going to keep support for hardware in forever. They've FINALLY begun removing driver support for hardware I used back in the 1990s. They seem to be de facto settling on a ~30 year support timeframe here (although, if there's any indication anyone is actually using any of this hardware, I imagine they'll keep those individual drivers.)
If anyone is seriously worried about being able to use, same, a 3c509 on a modern kernel (I used one of these back in the day), it *is* entirely possible to build kernel modules outside the kernel source tree. One could absolutely have an "ancient ethernet card" support package that builds these drivers, if there's truly any demand for it. Back when I ran a 3c509, I was also running a 386 then a 486 though. By the time I went socket 7 (AMD though), I was using RTL8139s -- essentially a 100mbps NE2000 clone. These suckers were VERY cheap, I recall I bought like a 5 pack for $25 LOL. Don't know how they ran in Windows but they were 100% fine in Linux. My recollection was a 3c905 (which I ended up with one of too at some point) had lower CPU utilization, but it wasn't high enough on these to be an issue.
So back in the early 2000s, I worked at the local university's surplus department, the university hospitals' computers also went through so we'd gt like 10,000 computers a year or so, and all the other equipment, we had movers and people to work on the computers (i.e. make sure drives were wiped and everything. We had a very nice tracking system to ensure no drive made it out of there without being 3-pass erased which involved both electronic tracking and physical stickers -- if a sticker fell off... which they did when it was very humid... pretend the drive is not wiped and start over.)
So we set up some pre-standard wifi devices we got (ran at something like 1 watt) and after several weeks all of a sudden, they would not even reach 20 feet. This went on for like a week. (We had it open to the public 1 day a week, the rest of the time was sorting all the crap^H^H^H^H prime vintage furniture and equipment that we brought in... plus give departments a chance to come in and buy some of this stuff back). So I'm out on the sales floor and an employee commented it seemed warm in this one corner of the floor. I walk toward this area and could hear a hum... yes, some jackass had powered up a microwave oven WITH THE DOOR OPEN and it had been running like that for a week. I snuck up to it from the side (so I wouldn't get microwaved) and unplugged it. The wifi devices then worked great again. What do you know, 2.4ghz wifi works poorly with a 1500 watt interference source LOL.
So glad I'm a Linux user.
Switched in 1993 to Slackware, used Gentoo for a long time and I've used Ubuntu since ~2010 or so?
Here's the COMPLETE list of upsells I've gotten from it:
1) During the graphical installer, when it shows those pictures telling you about how great Libreoffice etc. are, I'm pretty sure one of those screens mentioned extended security updates being available and so on. There is no clicking through extra screens to say "no", though, it's unobtrusive enough that I *think* there's one there but I'm not even 100% sure.
2) Ubuntu has a notification reminding to enable some kind of backups, and years back this 'set up some kind of backups' reminder mentioned they had a "Ubuntu One" cloud backup service, so it'd suggest you could use that. This was a one-time notification (and not full-screen, it popped up in the corner like a "your downloads are completed" or anything else.)
3) The update manager, and apt, show what updates you have available, but also shows what extra updates you'd get with Ubuntu Pro. (This has additional updates for packages outside 'main', and extends security updates for like 15 years instead of 5. In case one wanted to just keep running the same version of Ubuntu but still receive security fixes.)
4) npm (Node Package Manager) will ask for donations for some packages. That's true on any platform though, and npm didn't come out of the box on any of the distros I used.
That's literally it, and as you can see I'm even counting a 3rd party install in my list since there's just so few of them. Even the backup notification and apt/update manager one were HIGHLY controversial when they were put in, as was the change to npm. Linux users will truly not put up with this kind of thing unless it's minimally invasive (the reminder to backup -- which is still there, just not with a mention of Ubuntu One since I think the cloud backup service is gone -- you close the one time notification. And the rest, you can ignore the additional text and it doesn't affect the workflow of the app at all.) And even then it probably did drive SOME users away to Debian or Mint etc.
When AT&T wanted to buy out T-Mobile some years back, their inflated numbers on 'this will create jobs' not only included temporary jobs (people going out to towers to physically switch out hardware as needed), but also estimates on those workers like buying fast food therefore McDonalds etc. will hire more employees. (This merger got turned down partly based on a leaked document showing the unsurprising conclusion that ultimately the merger would cut jobs, since of course you merge two cell phone companies you will end up with peiople in both companies doing the same job.)
Honestly the current Mac Pro was kind of a joke. Same CPU as some Mac Studio model that was like half the price. RAM is soldered on so that's no advantage of the tower. It's got storage bays, but nothing to do with them since the storage is soldered on and no SATA or M.2 slots. The PCIe slots are there, but almost useless in macOS (since you need macOS ARM drivers and there just aren't that many of them. I will note, if you throw Asahi Linux on one of these, the PCIe is fully useable, since the Linux drivers are portable including to ARM systems. Even apparently the unnatural combinatoin of ARM CPU + Intel ARC GPU.) Apparently the one advantage of the tower is it's SLIGHTLY faster under extended full load operation (due to better thermals), but for the $1,000s in price difference once could have likely just pointed a fan at the Studio to make that up too.
To me this article is pure Apple fanboi'ism... like... Apple makes non-expandable systems, therefore everyone will switch to this.
There's those "Chromebook-based" systems where they've got some wheezy bit of RAM and SSD (sometimes a nice slow eMMC) soldered on, of course. But Intel has now announced "Lunar Lake" (which had RAM integrated on the CPU) is a one-off and they are not doing that any more, so that appeared to be a sign of this kind of trend, but Intels stepped back from doing this.. Anecdotal of course, but my mom got a notebok for $218 new recently (due to RAMpocalypse it now costs about $250). I really assumed (since one of the vaunted reasons for soldering on is to save the cost of the DIMM and M.2 slots) that it'd be a Chromebook-style board with everything soldered on. Nope! There's an access hatch at the bottom of the case, and it's using conventional DIMMs and M.2.
I don't see notebooks going all to "let's just solder it all on"; and desktops I DEFINITELY don't see this even starting to be a trend outside the Apple world other than perhaps the smallest systems (the ones that are basically a laptop motherboard just in a little brick-shaped case with no screen or keyboard rather than a flat 'case' with a keyboard and LCD built on). Any suggesting of SERVERS doing this is kind of a joke IMHO.
I do have a thunderbolt port or two on my desktop... but, I've never used it. Nothing against Thunderbolt, it IS amazing technlology. But, I added additional USB3 connectivity using a PCIe card, GPU via PCIe card. If I filled my M.2 slots I could add a card to add more M.2 storage too (although, to be honest, I have the vast majority of my storage in the form of spinning rust. I mean, 24TB of storage for maybe ~$400 over the last several years, even pre-RAMpocalyipse something like $2000-$3000 in SSD storage and now probably closer to the $4000-$6000 range.). In that vein I could also use those PCIe slots to pop more SATA ports for even more disk storage LOL.
Yeah the Mac Pro towers (the one made up through 2013) was impressive. Lots of RAM expandability, lots of storage expandability, and plenty of PCIe slots. The modern one looks like the old one, but the RAM is soldered, storage is soldered, it inexplicably has all these storage bays, but no way to actually hook them up to anything. And the PCIe slots, but there are VERY few PCIe ARM drivers for macOS. (But, if you put Asahi Linux on there, well, PCIe is PCIe so almost anything that'd work on any PC will work in there. Even the unnatural combination of ARM CPU + Intel ARC GPU.)
There are some that have done that. bcachefs is intentionally designed to have a bunch of advanced features, the likes of btrfs (but not so buggy) or ZFS.
At home I actually do use a layered approach -- I've got some data volumes with ext4 to store the actual data (since it's so good at it), and s3ql filesystem layered on top, providing deduplication and compression (which it does very well.) s3ql is designed to be able to store your data in a block object store, but also allows pointing it at a local directory and it stores the blocks in there instead, so that's what I do. Nice thing is it's *very* POSIX compliant -- I can run a VirtualBox VM off it, put the steam library on there, put source code in there and build out of it, etc....Also I do run filesystem backups of my other systems into one, the deduplication saves a lot there of course. Some of those fuse filesystems are pretty restricted, this one implements everything; it even supports extended attributes (but not ACLs -- access control lists -- I think for security reasons, they are actually filtered out in the code.)
Just had to second this. I've used btrfs several times in the past -- the first time I lost data. The second, third, and fourth times, the filesystem went read-only. No problem, I'll run fsck! No fsck. Every. Single. Time. this enthusiastic btrfs fan base insisted it was totally stable (as long as you didn't use a few odd bits of functionality like RAID-10, I was running a single disk so no problem) and I think the LONGEST I had it not blow up was about a week. IF you have perfectly reliable hardware, it might be reliable... but pull the power at the wrong moment and you're SOL. Oh I also ran it in a VM last year, since OpenSuse defaulted to it... after one unclean VM shutdown it blew out there too.
And yes, there was some kind of drama. It really looked like it amounted to the Linux FS devs being concerned bcachefs would be all unstable like btrfs, and they'd have to deal with the fallout for years, the numbers of devs at Redhat etc. working on this is now VERY small. The conversation got heated, there were even accusations of Kent bringing in people to the discussion to try to push things his way; to be fair, from what I saw the kernel devs were doing so to (getting people from outside the filesystem dev to throw down.) Ultimately, the final straw was making too extensive of code changes during the kernel release candidate phase.
Frankly, I don't have a side on this but I could see both sides point -- Kent's filesystem is well designed, and already in better shape than btrfs was for years (... and probably better shape than it is now.) On the other hand, throwing in big code patches during RC and all that, that just doesn't work.
Very interesting!
Two points...
A) I took some classes with Professor Jones, graduated in 2001. It's interesting to see he has a fairly complete treatise on trinary logic.
B) In case anyone wondered what a balanced trinary computer is... it has the values represent -1, 0, and 1. It's entirely possible (and Douglas Jones' paper discusses this) having the 3 values be 0, 1, and 2 instead.
