Windows 10 To Be A Free Upgrade: What Linux Users Need To Know

[TiberiusDuval]

Not to mention their latest partitioning defaults, where they have the boot partition separate. Yet primary. So on MBR systems that means pretty much no more slots left for multiboot. And it refuses to install into GPT if the machine is running BIOS, too.

It might get even worse in the future, as MS seems to be pushing "storage spaces" (LVM) lately (ReFS is only supposed to be used like that). I have no idea how (and whether) that works with multibooting Linux.

Yes, I had some nice experience when installing Linux for a friend. Who insisted leaving Windows too on his computer. Only one HD, and Windows has boot partition, system partition, and yes system restore partition. And off course MBR scheme. Very very nice. And off course all Windows partitions were primary partitions.

[Cyber Killer]

You don't need a primary partition for Linux, a logical is enough. All sold new PCs have got GPT, with that you can even boot with PARTUUID.

I've had many times problems with booting from a logical partition (it seems not all machines support this correctly), so I think it's better to have at least /boot as primary. Though I learned that it's best to prepare partitions for windows before installing it, that way you can have your own partition layout without microsofts insane ideas.

[Kano]

You made most likely a fault in your partition scheme. If you check "fdisk -l" and the first partition begins at 63 and not at 2048 (1 MB alignment) then this is wrong for all 4k hdds and ssds. Also you don't have got enough space to embed grub. Grub is not 440 bytes only, that was possible with lilo which works with blocklists and no file access. If you use mbr you do not even need a boot partition in the case you use lvm or mdadm (encryption is a different thing). Within that 1 MB minus 72 bytes the first grub stage can fit even with btrfs support. The other possible problem was a hard bios limit, like 8, 30,120 gb, 1, 2 TB depending on the time the board was produced. This is now never the case and can be fully ignored until your HD is bigger than 2 TB, then you should go for GPT+uefi.

[nanonyme]

I thought fdisk didn't understand GPT at all and just broke those. Need parted instead to edit GPT

[Zan Lynx]

I thought fdisk didn't understand GPT at all and just broke those. Need parted instead to edit GPT

I am pretty sure that this depends on your version of fdisk. I had thought the Fedora and RHEL versions had been modified to support GPT. I would have to check.

[Kano]

Newer fdisk versions support gpt, like the one from Debian jessie does.

[alcalde]

One of the biggest problems with Linux distros today is that they change too damn often. In other words, the support lifecycle for them is typically only about 1 year. And there's no upgrade path to the next version, outside of erase-n-reload. This is most unattractive, both for end users at home, and in a corporate desktop support environment. It's also most unattractive for commercial software vendors who may be considering a Linux version of their application.

I don't know which distros you're thinking of. I use OpenSUSE and that's supported for 3 releases, which is 8 months * 3 = 1.5 years. There's a volunteer effort that picks certain releases and doubles that, so you get 3 years on those.

Erase and reload? You must be thinking of Linux Mint, another of those distros that become a default recommendation despite major shortcomings. OpenSUSE can upgrade in place and has been able to do this for years now. Switch your repos to the new version, run "zypper upgrade" and you're all set. So much of what people label "Linux problems" often turned out to be distro-specific problems - generally Ubuntu problems, or in this case a derivative.

I use RHEL for precisely this reason - a long support lifecycle. RHEL6 was released in 2010 and is supported through 2020 - I've got 5 more years before I'm forced to upgrade. And that's a key point right there. Being forced to upgrade.

This is good for fire-and-forget servers where it'll be running the same software for eternity. It's not of any benefit to general desktop users. You end up running ancient, crusty software like Python 2.6 and then expect software vendors to support your long-dead libraries.

You're supposed to WANT to upgrade... you know, get improvements. You're forced to upgrade with Windows since you have to pay for it. You're forced to upgrade with proprietary drivers when vendors decide not to support your Windows version anymore just to get you to replace your perfectly good hardware. But I don't see how one can feel burdened by being "forced" to install shiny new software for free.

End users (whether business or home) don't give a crap about their OS. It's part of the machine in their mind and it doesn't add any value.

And yet it's of immense value. Linux distros are also far more than just the kernel anyway. I'm about to upgrade my new system to BtrFS - that's a major value to me. Similarly over the last four years that I've been using Linux full time I've seen a great deal of value, from improved package management speed to zram and zswap for my low-memory laptop (and until recently, desktop).

They have no reason to upgrade unless their old machine "dies" or they are forced into it for end of life reasons.

They have plenty of reasons. They just don't know what they are because they have little idea what they're doing and aren't paying attention. It's no different from those who ignore the "check engine" light on their car until it stops running, never defrag their NTFS disks, don't install updates, or my mother who never reboots her laptop and just ignores any pop-up or warning, even when it's flashing. That's not the fault of the OS.

Remember, people were *happy* with XP, despite it being a 13 years old OS.

People would be happy eating gruel, too, if they didn't know bacon-wrapped filet mignon existed.I was happy(ish) programming in Delphi (Object Pascal) too until I did a language survey for a start-up and discovered what I'd been missing out on. Often when people say "happy" they really mean "familiar with".

[Michael_S]

You're supposed to WANT to upgrade... you know, get improvements. You're forced to upgrade with Windows since you have to pay for it. You're forced to upgrade with proprietary drivers when vendors decide not to support your Windows version anymore just to get you to replace your perfectly good hardware. But I don't see how one can feel burdened by being "forced" to install shiny new software for free.

I agree with the rest of your points, but not this one. It's not that I dislike upgrades and enhancements - I love them. But the advantage a Windows users has is that most software he or she bought for Windows XP in 2001 will run fine on Windows 8.1 in 2015. That's not true for Linux - if I had an old install disk for a favorite program sitting around from 2004, I probably couldn't run it on Linux today unless it was statically linked to all of its dependencies, which is not common.

... and actually, this has made me wonder if the best way to write end user, non-high-performance software for Linux is to target the JVM or even the .NET framework (now that most of the .NET framework is open source). Then the version of glibc, etc... on the specific Linux install I want to run the program on doesn't matter.

[Luke_Wolf]

... and actually, this has made me wonder if the best way to write end user, non-high-performance software for Linux is to target the JVM or even the .NET framework (now that most of the .NET framework is open source). Then the version of glibc, etc... on the specific Linux install I want to run the program on doesn't matter.

Well... if you're doing proprietary software development... then in principle yes. You still have to carry your dependencies around with you, but that would help to limit your dependencies, so long as they're pure .NET or JVM dependencies.

[Michael_S]

Well... if you're doing proprietary software development... then in principle yes. You still have to carry your dependencies around with you, but that would help to limit your dependencies, so long as they're pure .NET or JVM dependencies.

Why not open source too?