Posts by amacater 121 publicly visible posts • joined 27 Jul 2014
For Confluence I can't understand why anyone would want it
There, FTFY
if you are a large organisation (say > 1000 employees) or constrained by regulatory pressure, how can you EVER migrate to cloud successfully?
For an individual, yes - having a small website / Wordpress instance hosted in cloud may cost less than running a machine at home - but not at scale.
Of the three or four hundred Linux distributions known to Distrowatch - 200 odd are minor derivatives of Debian or Ubuntu. It's not a million Linuxes, it's about five. Red Hat is definitely an also-ran with only Slackware being much older than Debian - there's a month between them.
Lovely - a new typeface and font. I assume accessibility studies have been conducted - it's appropriate for dyslexia, neurodiversity. It's available in every Latin language and for shaping glyphs in CJKV?
If not, go away again, Microsoft. If it's inappropriate for dyslexics and is a default, expect actions under s504 / s508 ADA at the very least. Have you run it past WCAG 2.1 AAA standards?
I have "proper" Red Hat around to check how it goes. A couple of times they've broken their update process - broken certificates, DNF that didn't. At least once they reverted the certificate without telling anybody particularly.
One RHEL release had a bug where you couldn't complete the initial install without contacting subscription servers which was impossible in the installer itself for that point release. I could demonstrate that it
was broken because the corresponding CentOS installed perfectly. As someone on self-support, I had the devil's own job actually reporting that one and being taken seriously
Quality control just isn't quite there even when the amount of covered software is small and updates (relatively) infrequent.
Just my €0.02
VMs and containers are the new frontier of things to bite you, I suspect. Chucking stuff into a VM, then regarding it as vital - now you've got another 10GB or whatever to snapshot and backup regularly.
VMs as cattle not pets - now you've got the problem of knowing *which* VM. Chucking it into the cloud and adding S3 storage or equivalent - did you remember to specify and pay for backup of that???
Creating an ephemeral VM - easy - I can do it 20 times a day. Maintaining a VM ...
And that's before you decide to base your ecosystem on Docker or similar: "I can't be bothered to build locally, I'll just pull down the latest from Dockerhub, sight unseen ... it'll be fine."
Likewise depending on GitHub repositories to be there ...
if you *are* relying on VMs - document the process for rebuilding them, go through and do that once in a while: archive the metadata in version control. For the stuff you care about, archive *all* the software that you need and build in SPDX or similar so that you know which bits depend on which, not just for your VM but also for software bill of materials ...
Debian and Cinnamon is a decent compromise - all the apps (and maintainers behind them) from Debian and a fairly vanilla Cinnamon. And no need for flatpak or snapd unless you want to install them.
As the universal operating system, our priorities are our users and Free software.
Non-free firmware *offered* as standard. This means, amongst other things, that WiFI stands a chance of working out of the box. if you're visually impaired, it means that the speech installer might speak to you on a modern Intel laptop. It's still being finalised - Debian Bookworm should hit another freeze round about today - but the way it's meant to happen is that the installer will use non-free firmware to get your Linux installed then offer you the option to uninstall any non-free firmware used in the install. This would leave you in the same position as having installed Debian using only fully free software.
It moves Debian to where Ubuntu and Red Hat and OpenSUSE have been for years, I think. It is an acknowledgement of reality that people need to be able to install Debian but it's not really an abandonment of Free Software ideals. We'd still much rather have fully free software - but after 30 years, it's getting harder not easier.
Note: This also makes it easier to divest yourself of non-free software other than firmware. You no longer have to potentially open your machine to all the non-free software offered in non-free, it is quite OK to have a /etc/apt/sources.list with just main and non-free-firmware for Debian 12 (Bookworm). See also https://wiki.debian.org/SourcesList
Running it here on three machines: I'd suggest that anyone who wants to tries out the Bookworm Alpha 2 .iso will find it at https://cdimage.debian.org/cdimage/bookworm_di_alpha2/amd64/iso-cd/debian-bookworm-DI-alpha2-amd64-netinst.iso
Full disclosure: I work in the team testing Debian media for each point release - this makes it easier rather than harder for real people to install Debian.
Windows 10 downgrade for the win - it's amazing how many business laptops were sold that way. Otherwise it's Linux all the way for me and for the savvy folk that I work with.
Office integration with email and calendar is the only reason MS stay in business - and even that is slipping now that you can't get Office so easily on a perpetual licence. Decouple apps from the underlying OS - as MS has done with Microsoft 365 - and you might as well run Linux everywhere. Nearly 30 years with Linux as my main driver here - and it's been the year of the Linux desktop since about 1995 - see Lars Wirzenius quoted in LWN.net and at https://blog.liw.fi/posts/2022/goalposts/
It would be lovely if three things were to happen:
* Raspberry Pi to upstream all their kernel mods and dtbs
* Pi hats of all kinds to work out how to interoperate with each other and work with other platforms - hats are great but tied entirely to Pi as an ecosystem
* Raspberry PiOS to actually talk to Debian about how to build an OS
As a Debian dev, I'm very biased but the first and last points are very important - Pi 4 and earlier models can (nearly) work with vanilla Debian but it is bondage and discipline - and there may be things that depend on a 32 bit armhf distribution that Raspberry Pi don't really care about.
SBCs from China and Armbian - don't even go there....
Nobody yet loves ARM _enough_ to build a supercomputer out of ARM cores - and even then, supercomputing is its own strange world.
If you've got a weather forecasting model cranking on PB of stored data, you need grunt processors to do the number crunching - it's not like VMs and webhosting where you can get away with some level of underprovisioning / overselling based on bandwidth used - _all_ your customers don't turn up at once :)
I'm a partisan - I've been running Debian for a *very* long time - but at least much of the non-free firmware hassle with Debian should go away relatively soon with the outcome of the Debian General Resolution to include non-free firmware in the installer as from the next major release.
Fedora includes non-free firmware by default. Ubuntu includes firmware - but not necessarily non-free firmware.
I like .deb - but it's definitely because of software curation and because a lot of people are working very hard indeed to make it work behind the scenes. There's not much to choose from .rpm and .deb - the quality of software curation makes a big difference and the long tail of Debian or Ubuntu derivatives have too few developers to keep them all going.
However successful the figures - this is a major corporate/cloud "thingie" - quarterly losses on that level are not wonderful. How much of that money is going into support burden / vulnerability remediation / building sustainable code for the future (and unifying product codebases and features as necessary) and how much is just loss?
So: if you head over to Debian mailing lists and see what the fuss is about. One of the maintainers of the Debian media has asked: Can we split off firmware into its own subset of non-free software - which stops you having to suggest the whole of non-free to users? If users want a fully working PC, can we then add the contents of the firmware repository to the install medium so that it works out of the box for most users?
This reflects most people's reality with installing Debian without firmware: it's really hard, stuff doesn't then work and people complain that Debian is at fault. Almost all the other distributions already accept firmware as absolutely necessary. The ones that don't - the FSF approved ones - are mostly three or more years old / unmaintained / are switching to a BSD base. Many - including myself - suggest installing with the firmware .iso file to save trouble you might have.
Our priorities are our users and free software: the discussion (and various options proposed) are as to how we best serve that. Both Steve McIntyre and Andy Simpkins have also written blogs on this subject syndicated to Planet Debian.
[Disclosure: I am a Debian developer: I work on the CD release team with both Steve and Andy].
This is a "We'll keep some of our data in our own data centre, thanks" moment for anyone looking to move to cloud offerings from Atlassian.
This is an "Instigate a two-person rule for major changes - and test, test and test again" scenario for anyone vaguely competent as a lessons learned.
This is "Keep your instances of production, dev and reference on the same version" scenario for anyone using Atlassian data centre versions at the moment - and probably a "distrust the sales droid from Atlassian who will try to upsell you cloud".
This is a "Run, don't walk, from a major purchase or future deployment of Atlassian" for many considering this.
Backups and DR don't solve everything - a pint for the poor Atlassian folk saddled with unfsck'ing the mess in spite of everything.
Technically, Ireland was united as one state for about a day, then NI opted out on 7th December 1921.
The most straightforward border to adopt would be down the Irish Sea and the Channel - bilateral
borders for England/Scotland/Wales with checks for all goods entering and exiting the EU. If Brexit means Brexit then it means hard maritime borders for all trade with the EU, surely.
As for fishing in the North Sea and Channel - that's been shared for at least 450 years.
So: Part of the problem is FF building requiring Rust - and thus effectively a new toolchain per distro per release - stable, oldstable, oldoldstable, potentially.
That also extends to building FF on non-Intel 32/64 bit versions but also all the other architectures that Debian supports. It's in hand: it will be done soonest.
There's a long thread in the debian-user mailing list at the moment: perhaps the most germane comment is https://lists.debian.org/debian-user/2021/12/msg00344.html from Roberto C. Sanchez.
Folk who have solved this by updating to FF95 themselves are finding issues with having to back up and restore profiles as new FF does things slightly differently.
As Roberto says, Ubuntu can afford to do things differently, not least because thy only have two primary architectures to worry about. There's also been a long comment thread on Phoronix which largely degenerated to general Debian bashing. Every distro has a trade off: the trade off of knowledgeable developers and the Firefox release cycle might be to push Firefox out eventually. It's actually very similar problems with chromium cross-distro when you dig into it. It will be a problem if we don't have distribution maintainers around for whoever are sorting out FF builds on a regular basis. [Full disclosure: I'm a Debian developer and help reply to questions on Debian-user but other than wanting a decent web browser, I've no great desire to participate in blame wars here.]
There's a carbon cost in transferring your data to and fro - it may not be much: it's cheaper for me to keep a local machine running to point machines at for Linux updates than it is to point 6 x machines at "the Net" - it's cer4tainly faster if it's on the local loop. Updates to that machine transfer minimum data - but there is the cost of keeping 4x disks spinning.
Paying for data transfer is the killer, I think, for bulk data. You pay once to transfer it and many times to access it. Silly example: Spotify used AWS then moved off and back to a cloud better under their own control. And I can't force a cloud provider to use power-effective hardware.
If you like Cockpit - nice sysadmin through the Web - it'll be lovely. If you like minimal change it'll be fine for the next five years - but RHEL 10 will be here in 2024 just as RHEL7 dies ...
If you want to do up to date software - meh, as always. If you've users who want to do stuff that isn't in core RHEL and demands third party repositories -- well, at least in RHEL 9 they will have sorted out some of the modules hell.
I've said it for a long time: support is a movable feast. Rather than migrate to "one of the Debian-based distros" - just use Debian (unless your bean counters want large scale support in which case pay for Ubuntu, maybe).
Tell me: Where can you get commercial training for CentOS? Commercial support? LPI will train for all Linux. To be honest, there's almost no difference between Red Hat and Debian now: same systemd, same apps, same desktop environment. [I've CentOS 8/ 8Stream / Rocky / Almalinux and Debian 11 on one set of VMs - put a user in front of any of them and you'd only notice that the Debian is more up to date and slightly more jazzy than Rocky]. RHCE is a course which demonstrates how to pass an exam.
Dell and HPE will support Ubuntu and Debian now: Lenovo are preinstalling other Linuxes - and second hand Thinkpads are the stock machines for savvy users.
Don't bother with the clones: come and talk to the Debian developers. Step forward six or so years from CentOS 7 to Debian 11. The reason there are 200 Debian derivatives rather than 15 or so for Red Hat?
Community - which Red Hat have rather squandered and fragmented in the last year or so. If Fedora is too fast pace - you've nothing. (Though the Fedora developers and users are a great community in themselves).
MATE - no problem - you don't even have to find a different install medium. www.debian.org will give you a download link for the netinstall medium.
There are three packaging solutons: tgz/git/build from source - Gentoo/Arch/LFS - .deb - Debian and Ubuntu and derivatives - .rpm - Fedora/RHEL and clones, and SUSE.
Apt wins dependency management for packages, I think but the others don't always have as many packages to care about anyway.
Of these for third party packages: .rpm is good for some third party commercial packages too but .deb and the Debian package universe has more or less won everything else.. [See, for example, the relative pain of producing a full bootstrap and rebuild of Rocky/Almalinux onto an ARM system like Raspberry Pi.]
Add in the horror of Red Hat subscription/license entitlement and .rpm is likely to die out in the long term, IMHO.
I would suggest you go over and subscribe at Beowulf.org. A tiny mailing list full of very bright people who've been doing supercomputing / HPC for 30 years or so. It was an accident that they started out with Red Hat packages to form Extreme Linux all those years ago: it's really not easy to build something from scratch but it is do-able. There are some supercomputing facilities running on Debian and Ubuntu, I think, and genomics sequencing is largely Ubuntu at the Wellcome Sanger Institute, for example.
Facebook are apparently using CentOS Streams - so on something that doesn't stabilise fully. What will be more interesting will be to see which CentOS fork Amazon picks as the basis for Amazon Linux which is the elephant in the room.[and the largest scale user of the Red Hat ecosystem.]
Fermilab and CERN were using Scientific and were mulling switching to CentOS after Scientific decided to not do CentOS 8 and to recommend CentOS 8. They're now, apparently, considering other options - one suggestion is that they might go to Ubuntu. Internal advice shows that a whole host of Linux distributions are made available internally.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
