For some reason I always assumed Google had their own customised OS built on top of Debian instead of using Ubuntu OOTB. Debian offers a much more stable base which someone with the resources like Google has can easily customise for internal Engg usecases. Esp. in the last few years, Ubuntu has been having a rough time with the ups and downs, while Debain-land has been much calmer. End of the day, Ubuntu has to bring in new bells and whistles for the user community (which has backfired multiple times) as it's targeted as a daily general-use distro while Debian only has to guarantee stability.
That's more or less the approach many Engg teams with much lesser Engg resources also take. Or are the Engg teams I've worked with just the minority?
The bells and whistles of Ubuntu includes the firmware and drivers necessary for computers to work out of the box.
In the past every time I tried out Debian on my laptops, I ended up with hardware problems.
Last time this happened, my Wifi wasn't working due to Debian not including the device's firmware. And there was no easy fix available, the only thing I could do was to recompile the kernel by myself. I then installed Ubuntu and it worked out of the box. Ubuntu is decent even on one of my MacBook Pro laptops I have, which is filled with components that need proprietary drivers.
My younger self wouldn't believe what I'm about to say, but I can't deal with hardware problems or any of that crap anymore. Still if I were to now buy a non-Apple computer, it's Ubuntu that I'd install on it.
On Google, they have people dealing with this shit, providing round the clock support and standardized hardware. That's not something that small companies can afford to do.
Often the required firmware is in the non-free section of Debian, which isn't enabled by default for idealism/purity reasons. It is very simple to enable installation of packages from the non-free section, and then usually everything just works. Basically, Ubuntu have chosen a different default - it probably doesn't actually have any firmware that Debian doesn't. Neither approach is wrong.
Or if you are installing from a CD or USB image you can instead grab one of the "non-free" images which should have the necessary firmware and repositories setup and ready to use after initial install. I don't think it's very easy to find these directly from the website but in case anyone's interested here's the location:
A handy workaround is to use your mobile + usb tethering.
Plug it in, then enable USB-tethering while logged into a wifi (data will go to the wifi not towards your mobile data plan then). The usb-ethernet driver works out of the box for almost all distros.
My younger self wouldn't believe what I'm about to say, but I can't deal with hardware problems or any of that crap anymore
Apropos of nothing much more than old people mumbling to each other about things, last week I bought a working, new PC (not peripherals or monitors), rather than assemble it myself from components bought separately from eBay for the first time in twenty years (well, 20 years ago I was using eSwap rather than eBay). Much the same as you; I'm just tired of putting it together and making sure to buy hardware that was all going to work together and then sorting out the drivers and all that sort of thing.
So I went to a pick-your-parts-and-we-assemble site, and bought a shiny new system with a Z370 chipset and a shiny new CoffeeLake processor. The chap I spoke to couldn't guarantee it would run a Linux, but he said he'd install Ubuntu on it himself and check it ran, before doing the usual Win install on it and sending it out. No worries, thought I. I'll swiftly turn that into a dual-boot system (which itself turned out to be a pain; this secure boot stuff sure can be awkward).
It turns out that Ubuntu 16.04 LTS doesn't come with a kernel recent enough to fully drive the Z370 and associated CPU inbuilt graphics, so I have to muck around getting hardware working anyway by bringing it forwards a few kernels, figure out how to add the relevent intel driver support to the kernel myself, buy a new graphics card and install it (no, I wanted to stop fiddling around with the hardware!)... or, as my older wiser self has decided, just keep using the old machine for graphically demanding Linux work and wait for 18.04 with kernel 4.15
It's been so long since I owned hardware made in the last six months that I'd completely forgotten about kernel support. I've been so happy for so many years that the kernel supports really old hardware, I forgot completely about the other end of the spectrum.
Imagine my surprise when I pulled a headless debootstrap chroot upgrade of a squeeze debian box that did not come up after rebooting.
I had to order a remote KMS session to figure out what went wrong: debian did not include the pretty common realtek-firmware and I was not able to apt-get it without a network interface. Luckily I was able to use it from a the previous installation.
Question: Is it Engg because it is suppressing the middle characters of the word, like Eng...g ?
That's really amusing from my perspective (having lived in an Asian country) because for us Westerns, most abbreviations come from using the first few characters of a word.
I've seen the same pattern in Japanese: The word "OSU", which people usually conjure to salute each other in the morning or as they enter a dojo or fight, is actually an abbreviation of "Ohayou Gozaimasu", built in the same way as I imagine the Indian English's "Engg" was built: Taking only the first and the last parts of the term. O(hayou gozaima)SU.
How many Googlers are using Linux for their everyday work? I assumed MacOS had as strong a hold on Google engineers as those anywhere else. I personally don't like using MacOS and would be thrilled to see more developers using open source systems.
Most use beefy desktops with Linux, and some laptop like a MacBook, Thinkpad, or Chromebook.
If you're working on a large project like Chrome or Gmail, anything less than 64gb of RAM and a giant SSD is a productivity hit, and most Mac desktops aren't price competitive with a beefy Linux workstation.
If you're in the market for a new workstation, HP's business workstations really are good machines (I had a nice one at a previous employer). They are nothing like HP's consumer PCs that most are familiar with.
Depends what you're optimising for? All rolling releases and recent stable/all testing releases run on roughly the same kernel, generally the same libc, with access to the same (non)free firmware if properly configured...
I've been very interested in exploring Chrome OS as a development machine even if the core system runs on a desktop like you said. Is ChromeOS good enough for this already? What about develop tools like editors/IDEs, etc ?
In my experience, Chrome OS is absolutely usable as a development machine for linux environments with crouton[1]. Crouton basically installs Ubuntu within Chrome OS, and can include GUIs and tons of software packages. I've done nearly the entirety of my WGU IT degree using Chrome OS, and if I get a software development degree, I'll use it there too. If Windows interoperability or virtualization is required, then you might look elsewhere, because virtualization is tough to set up in the chroot environment and would run sluggishly anyway if you use an ARM-based laptop like I do.
My understanding is that the software builds, runs and tests on desktops they SSH into. The sources either live on the Desktops or have a sync system setup. Either way, they need some proper editors or IDEs to work with locally on the CBs. Are there any tools with good enough experience that can do remote code editing available generally on any platform (and not just chromebooks?)
I would expect most of not all development would be done on desktops for that reason (and also for improved ergonomics and faster computers are always better) and there is no requirement for ide's to store any code locally.
As far as I know, most of us use Linux as main dev server, we are not allowed to put any prod code on laptop. So people just pick MBP and ssh to their dev servers.
This is what I would expect your PC or Laptop is should just a host for your X windows or Telnet / ssh sessions and any IDE's you use for code and sql.
This is very good news. Debian is still the base for a lot of the most known distros and is a great distro on it's own. Contribution into Debian Testing will end up not just in Debian, but in Ubuntu, Mint, etc. Debian doesn't have a RedHat behind it, like CentOS or Fedora, so a Google adoptation will certainly come as a boost.
Biased here, but at the scale of an organization like google why would they not use a tool like gentoo to customize and build binary packages for the whole organization? Does debian have symbols servers or something like that?
I think you've answered your own question here. An organisation the scale of google could use a tool like Gentoo to customize and build their binary packages. They haven't, and this perhaps speaks to the suitability of Gentoo as an operating system for use en masse.
Presumably individual engineers are still free to use Gentoo if appropriate but from a support and maintenance perspective the costs may not be feesible company wide.
It also matters to use something that everyone is more or less comfortable with. The more customizations you do, the more overheads you have to deal with.
Googlers must ask to use Windows because “Windows is harder because it has 'special' security problems so it requires high-level permission before someone can use it.” In addition, “Windows tools tend to be heavy and inflexible.”
You can still use an Android phone without the Google Play store if you use a custom rom such as LineageOS. I do not not have a Google account set on my Motorola E LTE.
Haha, good one. I have an Android device since last year (a dumb phone before then), so I am guilty of that one. I try to restrict use as much as possible (I have no Google account). I'll maybe try to install a Lineage OS on it, if that's possible (but afaik that's still Android so there's that).
Privacy wise, I am not sure if an IDevice would be such a bad decision. But afaik they're phoning home still and I have to trust them...
Windows 10 is the best OS in the family so far. The fact that you switched, is most likely just a hyped-overreaction to one of the things that they did that you didn't like.
It depends on your definition of "best". I agree with you from a technical point of view but this is not my sole decision criterion.
I had setup an Active Directory server on a Raspberry Pi for serving my Windows 7 clients. I would have loved to continue using Windows as I think it is great technology. However I was switching because of some crucial decisions by Microsoft:
1. Shoving unwanted "features" along with security updates. As much as I understand Microsoft in wanting a more uniform landscape this is a no-go for me. The only reason I was not hit by the Windows 10 mandatory upgrade was the fact I had the machines in an AD domain (or at least I suspect so).
2. Phoning home without any recourse for me to even know what is collected. How long it is stored. Remember: innocent data today could bring me in trouble tomorrow. Collection can change any time without me even noticing.
It's not necessarily a hyped overreaction if those things matter to the people switching. Some people actually care about those things that have changed enough to switch to a different OS that doesn't have those issues.
If you have a valid business reason for it, it's no big deal. Ie: I work with electronics EDA tools and the ones we use only work on Windows, so I have a laptop with Win 10 on it, but my daily driver is a Thinkpad with Linux.
I don't know how much they contributed to Ubuntu but if they also plan to contribute to Debian than it will eventually(matter of months) end up in Ubuntu.
I would think that a community supported Debian distro wouldn't match up to a canonical supported distribution. But it seems otherwise. The existence of Ubuntu has greatly benefited the Linux use base, with many patches contributed upstream to Debian and other related projects. Sad to see Google move away from them.
Personally, Ubuntu is much more palatable to end users who aren't CS majors, especially with their willingness to ship/support closed source drivers, unlike Debian
If they release gLinux to the outside world, we can all benefit from their constant vigilance, compared to just a community supported distro
And you already have that choice. What's missing is gLinux's availability to Google outsiders. I think gLinux vetted packages are likely to be more secure than the community edition.
But, if Google decides to monetize it, they may track the fuck out of consumers like they do on Android. That's the only downside I see to gLinux
I remember that Goobuntu was usually quite outdated as it is based on the latest Ubuntu LTS plus trailing behind a little. Googlers themselves were not tempted to use it at home. That does not matter for Google development, because all their tooling and code comes out of a monorepo.
Thus, I really wonder if gLinux wants to follow Debian testing closely? The might trail behind like the Ubuntu releases.
It wouldn't have made sense to release Goobuntu publicly, since it was simply Ubuntu with central management and logging, integration with internal account systems, maybe some hacky bits specific to their environment, etc. They've said roughly this much in past public talks about Goobuntu.
All of that makes sense for an internal corporate distro at a place like Google. It doesn't add meaningful value over Ubuntu for external users, except for state-sponsored attackers who want a more precise understanding of the vulnerabilities in Google's corporate workstations.
It seems like gLinux would have the same tradeoffs, except that they're going to push more of the broadly useful changes upstream.
(Disclosure: I worked at Google in the past, used Goobuntu, and am a Debian developer. But I was not on the Goobuntu team or involved with the decision to switch to a direct Debian base.)
I'd say the ease with which Debian can be kept up-to-date as a rolling distribution (just follow testing or unstable (which isn't all that unstable, breakage is rare, for that you want experimental)) versus the convoluted process of moving from one Ubuntu release to the next. Where Debian can be kept in line with apt update; apt dist-upgrade (or full-upgrade which does the same) Ubuntu generally necessitates a re-install between releases.
> Ubuntu generally necessitates a re-install between releases.
What? I have taken fossilized servers neglected for ten years and managed to successfully get them to 16.04 LTS - stuff is harder with Debian where e.g. mysql-server has disappeared in buster and replaced by MariaDB.
It is possible - but officially discouraged - to do a dist-upgrade on Ubuntu. Sometimes it works, other times it 'almost' works. It is the latter problem which makes the official discouragement a reasonable precaution against breakage. Things like the mysql->mariadb migration in Debian might necessitate manual intervention but this is made clear from the onset.
Although Ubuntu is a glorified copy of Debian, the components that Ubuntu changes tend to be brittle and unstable, and not adequately justified. If anyone wants an OS that helps them get stuff done, not mess with their work, and not even break between major upgrades, Debian is the way to go.
There was a time when Ubuntu had hugely better usability than Debian. Nowdays, Debian has mostly closed the gap. Thank you Canonical for the competition! Ubuntu/Canonical has lost its focus on the desktop. Desktops are boring now. Servers are more profitable and the entry to mobile has failed.
Maybe another window has opened for some desktop innovation now? I would try to copy stuff from Android. Snap [0] is a promising way to deploy apps (in general, self-contained stuff you cannot depend on), even commercial/proprietary ones. The "app store" is terrible, though. Androids intents might be worthwhile to copy to decouple things. Androids activity lifecycle might be interesting. All of this requires a large-scale long-term reengineering of deeper layers, which is practically impossible for hobbyists. You will get hit a lot by people for being not-Unixy (see systemd) and only after years of slog you might be able to show a superior desktop.
On the other hand, there is no money to be made with desktops, so why should a company approach such a risky project?
I believe it helped. Canonical invested in usability and raised the bar in general. Maybe for clarification, I'm thinking about stuff like the Papercuts initiative [0] in 2009.
There is a similar situation with clang and gcc. Clang improved error messages and raised the bar. Gcc follows. It would probably not have improved on its own.
That's only possible if you pay no attention to the hardware you buy. The last time I had to do any post-install config with a linux distro was when I installed kubuntu 5.10 and since Debian 4 onward I never experienced any problem.
Then of course companies like Google are immune to this sort of issue as they even design and sell linux hardware.
Ubuntu doesn't really have timely security updates for most of the packages they ship - those in "universe" etc are only randomly issued security patches, and you easily end up running unpatched stuff if you're not careful. With LTS releases you end up with having years of exposure to unpatched abitrary code execution bugs in security critical programs like firejail.
Because ubuntu is based on Debian, almost all software/drivers packages for ubuntu work on debian.
Using debian could maybe be compared to using vanilla JS instead of the shiny frameworks.
The downside is that, debian is very conservative in their releases. Most packages that ship with the distro or are in the repositories are a bit behind. I guess that's why Google switched to Debian Testing.
Btw even debian testing is comfortably stable.
For people who want to try debian, I advise them to try Linux Mint Debian Edition. It is a vanilla debian shipping with a good looking DE. I installed it on my mother's laptop, she's been using it for more than 2 years. It is so stable that, one time her laptop broke and we pulled her old netbook from storage. I only took the hdd from the old laptop and inserted in the netbook. After installing the wifi driver the machine has been running flawlessly for more than a year now.
I only do dist-update dist-upgrade from time to time, no further maintance was necessary.
I ditched Ubuntu in 2009, my reason was that it was buggy as hell. Usually a new release would fix few bugs and introduce new ones.
Actually even recently I got email that a bug that I was watching was closed because it was too old. I mean I stopped carrying about Ubuntu almost 10 years ago.
I've found that dist-upgrade in Ubuntu introduces lots of breakages in Ubuntu. Was workaround was to reinstall keeping /home untouched. Once though, because of some UI/UX problems with the installer, I ended up formatting most of my /home before I aborted the install. That was the turning point for me. I went back to Debian and haven't moved away since.
Opposite of my experience. Been upgrading from LTS to LTS in production for many years. Non-upgradable OS like Windows and Suse Linux used to warrant new server hardware, but now the servers live far too long since there is no need to migrate.
You said production so I'm assuming you are using server version.
The vast majority of issues I was having were around desktop functionality: connecting a monitor to laptop, weird issues with UI, wifi, putting laptop to sleep. Actually one of last issues that infuriated me was them being ok with release that that crashed (froze) every time you started a laptop with a proximity to 802.11n network.
Imagine trying to use your laptop to take notes in college and it froze few seconds after it booted up repeatedly. It took me a while to find out what the problem really was, then found the bug, realize that it was opened before the release, yet they still went with the release.
To make things even more frustrating at the same time they refused to include OpenOffice 3, because it might not be stable enough. Talk about priorities.
That was the time I started looking at alternatives, someone recommended OpenSuSE and I really liked it.
It seems Google is far big enough to handle its own Linux distro;
So it is very strange why they haven't done it since they even have created their own programing language.
Building their own distro from scratch doesn't make much sense. Why would they take on such a task when it's been done already, and they can simply customize a major one?
I'm surprised they didn't consider something like Manjaro Linux. I used Ubuntu for 10 years and just recently switched to Manjaro with i3 and I'm not looking back, it is that good.
Manjaro? The distribution that told users to change their clocks back on their OSs to get updates working? The one that last year couldn't patch Firefox for over 1 month because two package maintainers were on holidays, but Arch patched in one day? They can't even sort out bugs in package updater (recently leaking memory - 2.4GB RAM usage to update 120 packages?!). Last week I moved to openSUSE from Manjaro and it was really good decision, works better, it's faster, ex-Windowsers love it because of YaST2 [1] GUI manager for everything
If you have 10000+ machines, then you already have the infrastructure to building deb packages, and experience in running apt/dpkg based system, then the jump to Debian is much smaller than jumping to Manjaro/Arch.
There's a huge difference in you switching your single desktop to a different distribution and switching an organisation the size of Google. Going with Debian brings then closer to "upstream", while leaving them with a base installation that isn't much different.
I've tried Manjaro long ago, but these days i really like Debian. Never liked Ubuntu that much. I think Debian is really fast, small and easy to setup, but that's totally my perspective. I'm just a novice and still have to google lots of stuff. :)
What does Manjaro offer better than Debian if i may ask?
It's based on Arch so packages will always kept up to date.
I don't know about Debian, but the system will break every time I try to run do-release-upgrade on Ubuntu.
I've been using Manjaro for over a year now, never had any problem.
Pretty sure thats Ubuntu specific. I have machines here that run the "same" Debian installation for 10 years, doing dist-upgrades every few years and switching the disks (or the whole machine) underneath.
That's more or less the approach many Engg teams with much lesser Engg resources also take. Or are the Engg teams I've worked with just the minority?