Posts by williamyf

Windows 10 LTSC 2021 has five months before security updates cost extra

Microsoft never tried to be the "bestest". They are happy to be good enough.

If for win12 they dial back some of the privacy intrusions (totaly in their control), while some Main Linux distros ramp up their privacy intrusions* (which is not under Microsoft's control), then the gap may me narrow enough to not matter to the typical user.

Having said that, I use MacOS, so I have no bone in that fight. But Who knows What I'll be using in 2028 (when, par chance ALL my Hardware is due for a refresh, desktop, Laptop and NAS), let alone 2033. That's why I like to keep an eye on these things.

* Ejem cannonical, cof cof

I did not know LTSC 2021 had an ESU available. Nice!

This "slow death of old version" is nothing new.

The same palybook happened when people were asked to leave WinXP, and later on when they were asked to leave Win7. This will be no different.

By Jan 2032, when the last Win10 related updates cease, we will be long past the RAMpocalypse and NAND-Flashpocalypse .

We will actually be debating if we shall go from 10 to 12 or from 10 to 11, depending on hardware and such

Voyager 2 cheats the power budget for another year

I wonder

How long before NASA starts to cycle the scientific instruments. Turn one on for 6 months, verify it is operational, do science with it, turn it off, turn the next one on, verify, do science for 6 months, lather rinse repeat.

At least, with that cadence, we get snapshots of what is happening in Depp space.

Samsung warns memory crunch will last through 2028 as profit rises 19-fold

Re: Delay

107% agree. Nanya, who is the fifth largest behind CXMT (even though Nanya existed log Before CXMT) is expanding.

CXMT only dabbled in (LP)DDR , and is moving now into HMB.

WinBond got out of the race at the DD4 milestone, may try to enter the race again...

Plenty of reasons to cautiosly expand capacity, even at 19x profits

Veeam adds support for six more hypervisors in case you are up for a new one

Until HPE gets bored of virtualization and cloud AGAIN, or change licensing terms, like they do from time to time. Worked with their enterprise side from 1999 until 2004 both as customer and contractor, and keept an eye on them ever since...

If I were doing these "move out of VMWare changes, I'd move to something open-source based (either openstack or Xencloud) , even if it is the paid flavour of some company.

Easier to migrate from ubuntu openstack to some other openstack (like redhat openstack or mirantis openstack, or refference openstack), or migrate from citrix xencloud to XCP-ng Cloud than to migrate to a whole different (private) cloud solution.

And yes, I know the article is talking about hypervisors + advanced functions, and not (private) clouds, but in this day and age, if your new (private) workloads are not (private) cloud native, and you are not working to make your legacy (private) workloads (private) cloud native, "you are doing something wrong" tm

The majority of corporate IT is now off premises for the first time

Re: No problem

Hire me. I am 53 ¿Why you ask?.

Oh, I see. You are looking for the unicorn 21 years highly energetic highly dynamic manager with 15 years experience in DC operations.

Oh, sorry, not that, you are looking for someone, anyone at any age that has certifications on Windows, Linux, Cisco, Juniper, Arista, Servers, Storage, Electricity, Mechanics to repair the diesel generators, and HAVC systems expereience..

Say no more

After rewriting SQLite in Rust, Turso turns its sights on Postgres

You are 107% right.

The most interesting OS nowadays, from a kernel and new ideas PoV are QNX, Minix 3 and the L4 family of µKernels.

NT could also be interesting, if microsoft put the current semi-monolythic kernel on a diet and returned the kernel back to its µkernel roots, but we know that aint'-gonna happen.

AI storage boom is keeping Seagate's hard drives spinning

Re: Well that explains it

107% agree. You are correct. They are switching the factories towards the more modern/bigger drives.

But the criteria about which factories are retooled are very complex.

¿How compatible is said factory currently producing low capacity drives vis-à-vis Producing Mozaic 5 (or ePMR) drives (i.e. small modification vs big modification vs rip and replace)? For example, your 14GB Ironwolf is helium filled, and has 10 platters. The difference between that and a 40TB mozaic 5 drive is "only" different heads, different platters, and one extra platter. Meanwhile, a production lines of 4TB drives will have no helium handling, and probably 3 or 4 platters, which means the casings are different, the flis bearing motors are also different.

¿Can the factory be upgraded partially (i.e. only some production lines), or has to be upgraded as a whole?

¿Can it be upgraded while still producing HDDs, or do I have to keep the factory stoped while the upgrades are being done?

¿Do we have commitments to keep producing the older smaller drives? That applies to both customers (distributors, or companies like HP or EMC) to whom you sell, as well as providers (for example, these companies do not make their own platters, so, if you have an agreement in place to buy platters with low coercivity for a number of years, you have to fulfill them).

¿Do we want to hand over distributor/customer relations and market share in smaller drives to our competitors without a fight? Especilly the one that has neither HAMR nor ePMR (i.e. Toshiba), and therefore, is in a weaker position.

¿Do we even have enough HAMR heads (pun intended) (or ePMR heads) and platters with the required corecitivity to produce the big capacity HDDs if we reconvert the factory?

As you see, while seagate, WD and Tosh would love to reconvert every single factory, and churn out only max capacity HDDs, is not that simple...

A requiem for Optane, Intel's KV cache killer that could have eased the RAM price crunch

Agree 107% with you. And not just any old fab. We are talking here a VERY especialized fab, that would not work for anything else. Not for logic chips, not for DRAM, not for flash.

You forgot the joke alert.

Having said that, 2 months ago I went to my MBA 20 year reunion, and people from all over were complaing about inflation. A result of all the (needed and justified) COVID liquidity injection for stimulus and benefits/bonuses.

That is the explanation of the increase in price of plastic pots

Re: Software was a problem

In reality, in modern computing (and especially on servers, where optane was aimed) you do not want that.

Database or KV cache in memory space persisting over reboots? A-OK. OS persisiting over reboots? Almost useless. In this day and age, the main reason to reboot a server is to apply a patch, and that, by definition, means the OS can not persist. Servers are always on, even for most HW changes.

The "save a file" part is partialy true. You still want to save files to block storage to simplify backup and data retention tasks. But other than that, no need.

High hardware costs see Red Hat offer a two-server edge rig, no mini-PC required

Re: Anyone rember VMS clusters?

Many 2 server HA systems in the '90s used a couple of tricks. the first was an Heartbeat signal. a signal solely dedicated to indicate to one machine that the other was alive. Seen anything from 10Mbps thin coax to serical connections for that purpose. The second was a "token" file in the shared storage (if there was shared storage, and more often than not, there was).

The token was unclaimed when everythign was OK. Sometimes, the shared storage was runing certain FileSystemss with special tricks to help with this token thing.

If all communications (including Hearbeat) were lost between the machines, and one of the machines was master and the other was stand-by, the master one would:

1.) check itself.

2.) in case the watchdog decided it was OK, would try to get the token INMEDIATELY. IF the machine was not OK, it would inhibit itself.

3.) If token secured, keep working, if can not secure the token, stop working

Meanwhile, the stand-by would:

1.) Check itself.

2.) in case the watchdog decided it was OK, then WAIT A PREDERMINED TIME to try to get the token. IF the machine was not OK, it would inhibit itself.

3.) if you get the token, start the services and do work, if the token was not secured, keep doing nothing.

A nifty trick for extra security was that, if a shared SCSI storage shelf was used, the active machine was assigned the highest priority in the SCSI bus, while the stand-by machine was assigned the lowest priority of the SCSI bus.

IF the machines were both active (say, one was running services 1, 3 and 5 and the other was running services 2, 4 and 6) it would be more complicated. Dobuly so is 2 depended on 1, and 6 depended on 3 or somesuch.

Of course, if you had 3 or more machines in the cluster (for load distribution purposes) then normal quorum mechanisms applied, and you did not need all this HA and Token stuff.

A company called StorMagic, especializing in SMBs and "Large Distributed Companies" has a solution where the arbiter is a small VM on the cloud. Kind of a niffty idea if the remote location in question, even having local workloads, can not work without a connection to the company's private cloud.

A company in the USoA called Sheetz! (named after the founder's last name), operating gasoline stations with massive convenienece stores + restaurants attached, moved to this solution when running away from VMWare. They have around 800+ locations, each with a Dell 450 and a Dell 400 server in HA, and these arbiters.

https://arstechnica.com/information-technology/2026/07/sheetz-moves-838-stores-off-vmware-broadcom-created-too-much-uncertainty/

Again, is a way to lower HW costs. 800 distributed arbiters are a lot of arbiters to deploy and manage physically, with the associated costs. Centralizing and virtualizing all the arbiters lowers those costs.

The Document Foundation says Microsoft's file formats keep users on a short leash

I do not know how many people in this day and age create new (or even have) legacy documents with sections in wordperfect or smartsuite wordpro.

All the parts of OOXML pertaining new document creation are well documented. The only parts that are not are those pertaining legacy formats that microsoft had to reverse-engineer (like anyone else).

Are people really asking microsoft to develop a standard for renderig QuatroPro spreadsheets?

COME ON!

Comment in code read 'Dear future me, sorry I wrote this'

During uni, me and my coding partner were having a similar issue. We traced the problem to a specific line of code.

In desperation we erased the line and re-typed it from the book, and the error magically disapeared.

Our only explananation, to this day is that some non-visible character logdged itself somewhere, somehow, the firts time we wrote the line.

Mozilla speeds Firefox release schedule to biweekly

I settled on ESR for my personal use as soon as the first one was announced, and I do not regret the decistion one bit, even if it becomes a little painful in the last quarter of the life of the browser...

I follow carefully the whattrainarewein, and install the new one as soon as it lands. It is nice to have a modern browser for 4 weeks ;-)

On a more serious note, ESR is important for a few reasons:

1.) Many ISVs and SaaS platforms certify only against the ESR, not against normal release firefox, so, it is important what the new ESR brings to the table, as that is the feature set your are stuck with for a year if you use any of such ISVs SW or SaaS polatforms.

2.) As Liam alluded with the WaterFox example, the ESR is the basis of other important downstream projects. Another example of this is KaiOS. In this day and age not very relevant, but for a loooong while, a very important mobile OS in emerging markets.

3.) Many a Linux distro like Debian, RHEL Suse Leap, AntiX, Slackware, Kali and TAILS have the ESR, and not regular Firefox, as their default browser.

4.) Many an organization mandates ESR as their browser, regardless of the default that their OS ships with.

I have been anouncing the advent of ESR on ./ for a few years now. ¡How quaint of me!

Chinese memory ban would cut off RAMpocalypse relief

Literally, my second paragraph said that vast swates of the world that are not western companies will be the ones buying chinese RAM and NAND Flash

¿Remember the interview with the Steam Machine team when they said:

The memory makers (big three + 1) came and said: "This is the volume we are willing to sell to you, this is the price we charge, if you do not buy, do not talk to us again"?

https://youtu.be/3JElBQ3ooHY?t=327

Well, the Chinese companies help you with the "this is the volume we are willing to sell" part...

One would have tought that, after the replay you read the post thoroughly... but... i do not think that's the case.

Literaly, the first line of my post said that the chinese are NOT selling cheaper. just bringing more volume than what the bigunns are offering.

For what I have read on ElReg and elsewhere, YMTC and CXMTC are not selling cheaper than the competition, what they are bringing to the table is increased volumes compared to what their western counterparts are willing to offer.

But even if this restriction came to pass, I am certain that vast swates of the world will buy DRAM and NAND-Flash from these companies, no issues, Chinese companies themselves, SE Asian Companies, non-Oily middle easter states, vas swates of africa and LatAm...

That will, in turn, have the same effects as if American companies were able to buy their chips...

Windows 10 refuses to die, and the security bill is coming due

¿What's different from an LTS and Non-LTS version of Ubuntu that makes all other versions of Ubuntu imposible to support for 10 years?

¡Absolutely Blame Canonical for not supporting Non-LTS versions for 10 years!

I have no inherent sympaty (or dislike) for Microsoft (BTW I am typing this on a Mac), but let's be honest here:

If the company making the medical/industrial equipment (or somesuch) choose vanilla WinXP or Win7 instead of WinXP POS-Ready or Win7 POS-Ready, then the responsability (and stupidity) lie with the company for making that decision (their techies should know better).

If a company making medical/industrial or somesuch equipment used Vanilla Win10 instead of using Win10 IoT, again, the onus is on the company, not microsoft.

Microsoft publishes these support guidelines ~10 years in advance, and rarely change them, and when they do change them, is to extend the support even more.

For reference, Vanilla winXP support ended in Apr 2014, ESU lasted until Apr 2017 meanwhile Embeded was supported until Apr 2019 (The OS launched in 2001)

For win7, Vanilla support ended in Jan 2020, ESU ended in Jan 2021, while POS-Ready support ended in Oct 2023 (the OS launched in Oct 2009)

For Win10, Vanilla support ended oct 2025, ESU for consumers (first time it is offered to consumers) will end on oct 2027, ESU for bussiness will end on oct 2028, and IoT support will end on Jan 2032 (the OS launched in Jul 2015)

So, again, complain to the seller of the device for choosing the wrong version of Windows, not to Microsoft.

Is like a medical equipment maker using ubuntu non-LTS to make their device, and you complaining that the device was ""Outdated" solely because of marketing department of "Krapnonical."" And no, even if it is Linux, you are not free to go and change the underlying Linux from non-LTS to LTS (or to other Linux), lest you lose the certification, just like you are not allowed to change the Win10 in a medical equipment from Vanilla to IoT lest you lose the certification...

¿See the analogy?

Re: Cost versus risk

I agree 107% with you. Having said that, most modern-ish machines, even the lower cost ones, have NPUs nowadays. Now, if those machines wil have enough RAM to keep the models + the context + the OS + the APPs without trashing the Swap/PAge file is another issue entirely.

Let's also hope that MS also includes some usefull bumps in the requirments Like AVX 256 (instead of the current SSE4.2) and WDDM 2.4 instead of the current WDDM 2.0

Re: Which 16.9%?

«I was surprised when some mundane (ie. notechnical) friends actually told me that they knew about the Windows 10 security problems, so they do all their banking on their phones now. I hadn't considered that, but it's possible that the security issues might be mitigated somewhat by the younger generation who use their phones as the primary interface to the net, and see the PC as essentially just as gaming machine that shows Netflix.»

And then you realized, to your horror, that said phones were running Android 13

Re: Which 16.9%?

«There's also https://0patch.com/, which is better than nothing, but (a) isn't free, and (b) is questionable as to whether it's really able to properly address all security issues.»

There is no question about that at all. 0patch themselves are very honest about it, and in their own FAQ they tell you that they can not patch certain stuff. But, you have to dig deep into the documentation, and they try to downplay it... but it is there:

https://support.0patch.com/hc/en-us/articles/360011047979-Can-you-micropatch-all-types-of-vulnerabilities

«Unfortunately not.[...] We currently cannot directly* micropatch:

Vulnerabilities in Windows kernel; this is due to the fact that Windows have an active self-defense mechanism for preventing kernel code modification called PatchGuard (formally named Kernel Patch Protection), which our micropatches would trigger and cause the computer to reboot. We believe PatchGuard adds negligible security value and there are known ways of disabling/bypassing it, but we have for now decided not to do that. It is possible that we'll revisit this decision in the future and advise users to disable PatchGuard (e.g., for end-of-support Windows 7 and Windows Server 2008 R2 servers with 0patch installed) in order to apply kernel micropatches, should these become necessary.

Vulnerabilities in .NET code; .NET applications are not compiled to native machine code but to managed code, which gets interpreted by .NET Framework's Common Language Runtime upon execution. We can, however, micropatch .NET Framework's native executables.

Vulnerabilities in Java code; while we can micropatch the Java Runtime Interpreter (e.g., java.exe), we can't micropatch Java applications (typically delivered as class files in a JAR archive) because they, too, are not in native machine code.

Vulnerabilities in scripted code; it would be relatively simple, technically, to modify scripted code in memory right before it got executed, but in contrast to binary files, which practically never get manually modified, scripted code is often modified by users or admins - and that makes it hard to identify a vulnerable version and pinpoint the vulnerability within. This is why we decided not to support micropatching of scripted code.

»

Re: Which 16.9%?

Lansweeper is mostly an enterprise tool. So, their numbers are mostly for enterprises.

Starcounter was also mentioned, but that is a mixed bag.

What one deeds to do is not to look at individual numbers, but at the trend over a 4~6 month period, or even longer.

How to teach an old Intel Mac new tricks with OpenCore Legacy Patcher

The utility diminishes significantly because after (boreal) autum 2028 ALL intel macs stop receiving macOS security patches. Be them on OCLP or not. Does not matter if your machine was one of the only 4 models that get Tahoe naturay, or the OCLP team did the deed and cracked the Tahoe puzzlew, in Boreal autum 2028 there will be no more seurity patches for intel Macs...

So, very restricted webrowsing and internet activity significantly diminishes the utility of a machine, at least, for me.

But I am not your mom, I am not your daf, if you wan tot download videos of tentacles eating k0rn from questionable sites using a machine that has gone a copuple of years without security patches, you do you.

The project is progressing slowly, because the task at hand (taming the T2 chip) is hard.

Having said that, Tahoe is the last intel MacOS, and that will stop receiving updates in (boreal) autum 2028, so, after that, the usefullness of OCPL, and an intel Mac, diminishe significantly, unless you install T2 Linux, or put some Windows Server with Desktop Experience (2019 or 2022) on the bootcamp partitiuon to keep security updates coming.

Re: Wish this was around years ago

While is true that OCLP was not around before ~ Dec 2020, before OCLP, there were dosdude1 patchers. And even before that, there were MacPostFactor and MLPostFactor, to do exactly that (install newer versions of MacOS on Mac hardware that offciially did not support it).

MacPostFactor came about in ~2013 ( https://osxdaily.com/2013/04/20/mlpostfactor-installs-os-x-mountain-lion-old-unsupported-macs/ )

and Dosdude1 patchers came about in ~ 2016 ( https://www.youtube.com/watch?v=2eMGd3ivGQ0&t=1 )

So, plenty of options to breathe new "MacOS" life into your old Macs at that era without going to Windows.

A few points about OCLP

There is a ßeta of OCLP 3.0 that does run Tahoe, but it needs very specific circumstances:

1.) No T2 chip. The machine can have a T1, but T2 is a big no-no

2.) Due to changes in background processes in Tahoe, including but not limited to Aqua, the baseline processor usage is significatively higher than in Sequoia, so, if you want to update a (very old) Mac sans T2 chip, is has to be some sort of top of the line unit, lest the expereince be 5h17e.

Since Sequoia will be supported with security patches until (boreal) autum 2027, the Fine people from OCPL have that leeway to try and make a breaktrhough on the T2 Situation.

Also, please notice that, if you are doing OCLP today, it does not make any sense to install something other than Sequoia, as Tahole will lead to headaches and instablity, while Sonoma will lose security patches in a very few months (less than 6). And anything older than sequoia if not receiving security patches for years now...

People that install unsuported OSs say "no worries, the machine will not be connecte to the net", 2 weeks later, you see the machine is connected to the net. Then they say "do not worry, I use safe practices when browsing" some weeks later, when the machine borks because of a virus and such, and they ask you for help, you discover terabytes of tentecles eating K0rn videos are on the storage... Safe browsing practices my A55. So no, do not delude yourselves, install an OS that will receive security patches and get done with it.

As for Linux on Macs, is not as straightforward as installing Linux on a run of the mill x-86 Windows PC. In Particular, if your Mac has a T2 chip, you may need to take a look at the T2-Linux project ( https://t2linux.org/ ), and, while it can be aplied to any distro, is recomended for minimum fuss to use only the Distros they recommend...

Good luck

PS: My early 2025 maxed out Air is running Sequoia on OCLP. Meanwhile, I am waiting for Tahoe support for it, and for my 2018 Maxed out Mac Mini. Let's see what happens by (boreal) autum 2027

RISC-V firmware project wants every board booting from the same hymn sheet

UEFI is free (as in beer) to implement. IF what you want to do with UEFI does not use any licensed IP, you pay 0,oo £€¢¥$

If what you want to implement requires someone else's IP, you can license that via FRAND.

So, if one is creative enough, one can implement an UEFI firmware, without using anyone's IP. Not only that, any propiertary IP will be free (as in freedom AND beer) after 17 years from patenting (kinda like what happened with MP3 a while ago). So ¿Why exactly deviate from UEFI, which is already supported on RISC-V, and is the most popular MODERN firmware in the world?

Frame: A new X11 server – implemented directly in assembly

gripes with vive coded assembly

At some point the vive coded program will fail. Al all programs do, including human coded ones.

Is already hard enough to debug/modify/maintain software vivecoded in "normal" language. I can not even imagine how things will get with vive-coded assembly.

As a fun project, ok. As a serious tool... No way on earth

Microsoft emails Windows 10 holdouts: Fine, keep your old PC another year

Re: They are going to have no choice but to keep extending this

In less wealthy countries (and I know first hand, I live in venezuela and worked in colombia), when RAM and HDD prices were low, there was no way to make people move from XP to Vista/7. people hacked the OS to get ESU and POS-Ready (that era's IoT), and even server patches, even if that made the OS a vulnerable frankenmoster, instead of moving into the new OSs

But that was not only due to lack of funds, it was resistance to change.

Some people saw staying on an unsupported system as some sort of "badge of honour". I saw it with 98->2000, 98 -> XP, XP-> Vista, XP-> 7, 7 -> 10, and now we see it with 10 -> 11... If you do not believe me, do some paleontology in the wayback machine on forums of the era...

Nothing new under the sun.

Re: Family IT support...

While I agre with the recomendation of Windows Server 2022 (with desktop experience), WS2025, (with desktop experience), may be a bridge too far for (drivers wise) a machine that can not upgrade to Win11.

Remember, much of this "Install Windows Server with Desktop Experience to extend support" depends on WinServer being happy with the drivers for such old hardware.

And also with your SW being happy running there.

Also, may I point out that WinServer 2019DE, supported until Jan 2030 is also a cromulent option if WS2022DE does not support the drivers of your old hardware, or if your software is not happy running there..

Also, may I point out that Microsoft allows for a trial of WinServer (Example for 2022: https://info.microsoft.com/ww-landing-windows-server-2022.html ), 90 days if memory serves, so you could install both, and see which one works best in your old hardware...

Unlike Win10 LTSC IoT, which puts you in a gray area from a certification, regulatory, legal and EULA areas, individuals can buy WinServer DE witout impediment, so, if you care about such stuff, WinServer DE 2019/2022 is the way to go.

Re: Smug Win10 IoT LTSC user here

Please remember that the Windows under ESU is 22H2 (with some stuff backported from Win11 after that date).

Meanwhile, IoT LTSC is 21H2. which means that, while some of the userland stuff may be commonal, and therefore, patches developed for IoT may run on ESU too, Kernel stuff is potentially different between one and the other.

Even if Microsoft suddenly felt generous and allowed us to use 21H2, I am not sure if you could do an in-place reversion from normal 22H2 to IoT 21H2, instead of a re-install.

And let's not talk about what would ISVs (like game developers and Office suite and Image/video/sound manipulation SW vendors) say of such a change.

So yes, the existance of IoT LTSC 21H2 does help us on ESU, it is not a panacea, as it does not cover the Whole of the ESU/Normal OS.

Lenovo denies using banned Chinese SSDs where they're not allowed

Re: Beijing encourages Chinese companies to buy from their local peers.

Depends on the company. Large companies on the international stage and small and medium companies doing intl bussines and not under sanctions will not be be encouraged as much.

Ubuntu emphasizes Arm64 support – and gets Rustier

On your homelab, and SMB, sure, servers can easily reboot.

More so nowadays, now that systemd, and other advanced init systems, decreased SIGNIFICANTLY the ammount of time needed to shut down and initialize systems that are rebooting compoared to sys-v-init.

Move to large companies, or companies handling massive ammounts of incomming/outgoing traffic, or OLTP galore, or HPC, and things are not so rossy for restarts...

AMD’s Ryzen AI Halo makes local AI look easy, but at $4K, easy doesn't come cheap

CLustering is a diferentiationg factor between AMD and nVIDIA

IIRC, nVIDIA's DGX Sparc can go to only 2x clustering for a total of 256GB RAM in-cluster. Meanwhile, AMD can scale to 4x clusters, for a total of 512GB RAM.

If your workloads demand this size of RAM right now, or if you foresee you might during the lifecycle of the system, then AMD it is.

PS: Apple can also do 4x clustering but they have machines with 256GB, so total cluster RAM is 1024GB of RAM.

Re: What about the important question...

Yes, both classic and remastered, but not on stunning detail, because of the old-ish RDNA 3.5

T-Mobile appears to be quitting VMware – and fighting a very familiar battle for support rights on the way out

Re: Meanwhile…

OpenStack technical trainer here:

If you think of VMWare as virtualization only solution, like we still are in 2006, then sure, KVM is a cromulent option.

But nowadays, VMWare, ZenServer et al are used to make Private Clouds, or the fleets running on them use a few advanced functions beyond virtualization, with all that implies. Very few workloads are virtualization only, not touching any of the advanced or the cloud-dy functions .

The linux equivalent would be OpenStack, with all the load that implies.

There is another aspect in this too, and it is Application support. Many ISVs certify their platforms/apps on specific OSs/Distros running on Specific Hypervisors.

So, for instance, your ISV may say: Only Windows Server 2022 or 2025 only, RHEL 10, or Suse 16, on top of VMWare, Openstack or azure.

And there you are, for those workloads, you can forget about all the other solutions (obscure or not) that homelabbers love to peddle. Big corpos can pressure smaller ISVs to support their preferred solution, but the big ISVs will most likely put a few options on offer, and that's it.

In those cases, large intitutions (like T-Mobile, the focus of the article) have 100s or even 1000s of ISVs some more crititcal than others, and they need to reach commonality of solutions, or personel requirements ballon (the legacy VMWare group, the Openstack group, the xenserve group, the proxmox group, the azure group, the nutanix group, the....) along with all the other support functions (negotiations and keeping track of support contracts for each technology). A veritable nightmare. So, unlike homelabbers, Big corpos will probably go to a one or two vendor solution for their internal clouds.

Since VMWare was the leader, and for many lustres a model citizen, pretty muche every single ISV offered them as a supported option, therfore, it was the easier default.

Re: we'll see you in court

That 100-200Mbps is either a typo, or the fraction of the 40Gbps pipes that is available for the migration while you keep the current services running. Is not like you will stop all services while you migrate.

Mageia 10 keeps the 32-bit Linux flame alive

PErsonal diferences and politics sometimes lead to technical differences. For example, Mageia supports 32 bit linux due to personal taste and internal politics, meanwhile, openMandriva does not. yet this is (a very big) technical difference.

The more they reamin separated, the more they diverge, and the harder it is to re-merge them.

Re: There is no such thing as 32-bit and 64-bit

According to the Mageia 10 relase notes, it requires SSE2

https://wiki.mageia.org/en/Mageia_10_Release_Notes#32_Bit_Support

So, pentium4 onwards. With Via and AMD, the situation is murkier. See here:

https://en.wikipedia.org/wiki/SSE2#CPU_support

Lets hope that, if Mageia 11 keeps support for IA-32, they require at least SSE3, which provides a performance boost over SSE2 in workloads that use SSE, like FFMPEG (and, by extension VLC), Handbrake and image/video processing programs (i.e. FOSS equivalents to premiere pro and photoshop). SSE4 would not make any sense in 32 bit Linux land, as, at that point, all procesors supporting it were 64 Bit anyway.

An SSE4 requirement would have made a HUGE difference for 32 bit Windows* (were sometimes you needed to use 32 bit windows on 64 bit processors due to closed source 32 bit only drivers) as SSE4 accelerated other workloads beyond multimedia, but in Linux, with most drivers being FOSS (either by the HW manufacturer or reverse engineered), if the CPU is 64 bit, is 99.99999% sure you can run 64 bit linux.

By the way, SSE/AVX requirements are a nice way for OS and App makers to wane CPU support, that is unsderstable by the layman, deep enough for the tech person, and brings performance improvements to the table. If Microsoft were wise, they would have demanded AVX2 support with Win11 on top of all the other requirements**, and saved themselves the complains of Core2 and 1st to 3rd gen Core i people, while bringing actual perofrmance improvements to the table due to the new baseline.

-----

* Alas, every single 32bit verison of windows from Vista onwards demanded SSE2 only, MS never had the guts to gut (redundacy intended) processors during the Vista to 10 period. That's why my lowly T2080 Toshiba (with an Xpress200 chipset) could go all the way from Vista to 10 (with maxed out 2GB RAM, SATA SSD, and software overclock of processor and Graphics core). By the way, that machine is now in Cienfuegos, Cuba, still running Win10 and still doing good work (when they have electricity) thanks to the extended Windows ESU (it has an AntiX partition, by the way, which the new owner does not use)

** For some weird reason MS explicitly calls for SSE4.2 in the requirements page of Win11, even though all processors in the supported list support AVX2

From time to time, 23-Bit machines fall into my lap (pun intended). Some are of the P4 variaty, some are of the Pentium-M variety, and some are of the Core Duo (sans '2') persuation. For some people in Venezuela, (and some of the cubans from the medical mission), those machines are decent enough at least to write letters and browse eGovt and banking sites.

I tend to refurbish those machines with either AnitX (for the slower older ones), or Mageia (for the more capable ones). Mageia in particular is important, as OotB it looks more like Windows than AntiX OotB. So, for me is good that Mageia 10 is out of ßeta.

What I find strange is that for 32bit the Mageia team choose Kernel 6.18 (an LTS, supported until 2029) instead of Kernel 6.12 (a CIP SLTS supported until 2035). One would think that niche distros should make it as easy as possible for themselves to get the maximum external help possible, and getting security patches for the kernel from elsewhere means less internal work. But then again, I am not part of the Distro team...

I actually came to know that Mageia planned to keep 32Bit support thanks to Liam, so there is that.

But if your machine is of the 64bit variety, for all that is holly, use a 64bit Linux OS. Using a 32Bit distro on a 64 Bit capable machine comes from an obsolete belief that 32Bit will be faster because executables will be smaller, and fit better into caches... The extra size of executables is nothing in comparison with the extra General Purpose Registers and more eficient instructions that 64Bit OSs and Executables opens up. Especially in 2026, with many decades of compiler optimizations applying their full power on FOSS codebases (i.e. you are NOT running a 64Bit executable compiled 24 years ago, the executables where compiled like less than one year ago).

The only reason HW under linux needs a 32Bit OS, is if it has hardware that has only propiertary 32Bit drivers and no FOSS ones. As long as there are FOSS drivers (either provided by the manufacturer or reverse engineered), you can run a 64 Bit version of the OS.

No more Java refills for Intel Macs after JDK 27, says Oracle

Re: The impact will only be felt in late September 2027...

If your Intel Mac has a T2 chip, installing linux may be tricky. You will need a set of patches and scripts called T2 Linux. While it can be applied to any distro, is better to stick with recomended ones, and even then, compatibility is not perfect.

I am sticking with OCLP for the time being, then will donate the machines (in [boreal] autum 2028) and let the new owners decide what to do, and implement said decition.

Unless you code for JDK, or your 2015 Mac has a T2 chip, chances are good that you can get the party going thanks to OCLP (OpenCore Legacy Patcher).

OCLP can run MacOS 26 (a.k.a Tahoe) on macs without a T2 with reletive ease (and your 2015 MacPro probably has enough grunt for the task), and MacOS 26 will receive security patches until (Boreal) Autum 2028. But, until Autum 2027, please stay put on MacOS Sequoia, and let the OCLP team iron out the Kinks of Tahoe support.

As for the JDK, maybe swtiching to the OpenJDK can be an option if support there is better than Oracle's

Re: Rosetta

The Rosetta 2 Plumbing will remain for certain software. Mostly videogames, and maybe some other critical infrastructure. Perhaps some enterprising programmers will use that to re-inject broad Intel App support. ¿Who knows?

We may call that effort OpenCore Legacy Rosetta, or OCPR...

Re: The impact will only be felt in late September 2027...

Actually, no, the impact will be felt in (boreal) autum 2028. MacOS Tahoe (i.e. MacOS26) is the last MacOS with Intel Support, and it will receive security patches until (boreal) Autum 2028.

Having said that, OCLP is currently having SERIOUS troubles making MacOS26 work on Unsupported Macs with T2 Chips. Meanwhile, unsuported Macs with no T2 chips and Hackintoshes can run it with comparatively much less troubles.

But yes, as expected, some ISVs are ramping down support for core Apps in anticipation of the end of the Intel Mac era.

I do not develop in Java, but I need the Java VM for JeD (which I use as a Visio replacement). Lets'see how that goes.

BT and Verizon spin off international networking arms into $4B joint venture

As a former BT International employee (with Comsat in LatAm), i'd say this is good for customers in all Geographies, and for employees in areas where there is no overlap.

Customers in areas with overlap will enjoy a more diversified pool of engineers and the better of the two infrastructures (or even better, a more diversified combined one). And, in areas where there is no overlap, some multinational customers suddenly can call their prefered provider in places where, previously, they had to "fend for themselves".

Employees in areas with no overlap will enjoy increased bussisness (from all the multinational customers suddenly knocking on their door that could not be served before*), and therefore, more job stability, and more money incoming to fund projects that before were on hold.

Sadly, in places where there is overlap, people, specially in support functions (Sales, Marketing, HR, Fiannces), will be fired. And also, there will be knock-down efects on companies/subcontactors that served said companies, as everything is "rationalized". All the best to them.

* For instance, multinational customers of BT operating in a country where BT Intl. had no precence, but that Verizon Intl was serving, and Vice-versa. In those cases, HQ of the Clients will pressure the local office to move to the JV as soon as it is feasible, for economies of scale, negotiating power and "uniformity" of service.

O2 joins UK 2G switch-off with summer 2029 start date

Re: 5G is poorer when it comes to building penetration

Penetration and range have very little to do with the G, and more to do with the frequency band.

If you put 2G (GSM/GPRS/EDGE) and 5G in the same sliver of 900Mhz spectrum (say, 40Mhz each), range @ same power level, and penetration @ same power level will be very close. Probably better in 5G than in 2G because of the extra FEC (Forward Error Correction) of the more modern protocol (ditto for 4G), so there is that.

And before anyone says anything incorrect: The speed of 5G on a 30Mhz channel in the 900Mhz band will be the same as the speed of 5G on a 30Mhz channel in the 40Ghz band, as the modulation is the same

The reason you get worse coverage and penetration for 4G and 5G nowadays is threefold:

1.) The spectrum with the best propagation and penetration characteristics is currently being used by legacy 2G and 3G (and UHF TV)

2.) 4G and 5G tend to use higher frequency spectrum because there you can get bigger contiguous chucks of spectrum (say, 100~200Mhz) which is what those technologies crave, meanwhile, in the sub 1Ghz band, chucks were 30~40Mhz if you where lucky, ussualy less.

3.) Since the advent of 1800/1900 GSM, the handsets always check from higher frequency to lower frequency, and latch on the very first usable one they catch, which may not be optimal (you may want to go 1 or 2 frequencies below before latching).

Maybe in 6G we can bake in the standard that the handsets latch onto at least one narrow low frequency channel for things like notifications and such, and latch onto a wider high frequency one for data heavy usage. (multichannel has been a thing since 4G)

IBM stacks up a sub-nanometer chip future

Re: IBM Research - why ?

Rapidus Foundry and Samsung Foundry are licensing the Nanosheet technology today, and paying good £€¢¥$ for that. Samsung is actually producing chips with it, as of now.

Probably they will also License this NanoStack technology to some foundries (probably those two, again), and again, IBM will charge good £€¢¥$ for it...

So, is not only "Halo effect of obscure papers".

Not only that, IBM needs some specific stuff in the foundry process for their Power9 and up processors, so this research allows them to insert that into foundry partners. Back in the day there was a kerfulkle between IBM and GloFo about this... Ask Gemini about it ;-) if you want more details.