Talk:Installation guide

From ArchWiki
Jump to navigation Jump to search

Read this first before adding new suggestions

  • systemd tools such as hostnamectl, timedatectl and localectl do not work in the installation chroot environment, so please do not propose to use them in the guide unless you can prove that they have been made to work also in that case. See [1], [2], [3] and [4] for some past discussions about this issue.
  • localectl list-keymaps does not work due to bug FS#46725. For the chosen replacement command, see [5].
  • localhost must be set explicitely in /etc/hosts, as it is otherwise resolved over the network. See FS#56684.

-- The ArchWiki Administrators 22:17, 2 September 2016 (UTC)

Link to the German version

Instead of de:Arch Install Scripts you could choose de:Anleitung für Einsteiger it means "Beginner's Guid" and is a very detailed artikel for very new arch users and the future experts.

Thank you, done. -- Kynikos (talk) 16:31, 6 February 2018 (UTC)
This was already proposed last year and rejected: [6]. I don't see what has changed since then. If someone adds me as admin to the german wiki or changes the protection settings, I can update de:Arch Install Scripts as required. -- Alad (talk) 18:13, 6 February 2018 (UTC)
I see, I didn't remember that discussion so I've reverted the change, hopefully you'll make it to update the translation, let's leave this open until the problem is solved, otherwise this kind of suggestion will keep appearing recurrently. -- Kynikos (talk) 17:53, 7 February 2018 (UTC)

Network configuration

dhcpcd for installed environment

How about we add a sentence to Installation_guide#Network_configuration describing how to restore use of dhcpcd as done on the installation medium? That way if the connection already worked on installation people only need a brief look at dhcpcd. -- Alad (talk) 15:48, 27 May 2018 (UTC)

I think that would just hide the possibility of choice from users. Maybe just say that dhcpcd is not enabled on the installed system without recommending anything. -- Lahwaacz (talk) 17:34, 27 May 2018 (UTC)

once chrooted, one has to configure again

One chrooted, configuration file edited in the previous steps, will have to be redone in the chrooted environment (/etc/systemd/network/ files). This is the same for ntp configuration (/etc/systemd/timesyncd.conf).

This should be worth noting.

—This unsigned comment is by Mrechte (talk) 12:27, 14 May 2019‎. Please sign your posts with ~~~~!

The guide does not say to edit anything under /etc/systemd/ in the first place, so there is nothing to be redone. Furthermore, Installation guide#Network configuration says to "Complete the network configuration for the newly installed environment." -- Lahwaacz (talk) 17:27, 17 May 2019 (UTC)

Why should a static IP be preferred over 127.0.1.1 in /etc/hosts?

"If the system has a permanent IP address, it should be used instead of 127.0.1.1."

I think the ArchWiki should not just say do X but also why. Alad as you added this, perhaps you can explain?--Larivact (talk) 15:14, 21 May 2018 (UTC)

In Network_configuration#Local hostname resolution: "For a system with a permanent IP address, that permanent IP address should be used instead of 127.0.1.1." -- Lahwaacz (talk) 06:48, 22 May 2018 (UTC)
First appearance in our wiki, cited source, also discussion. -- Kynikos (talk) 10:26, 22 May 2018 (UTC)

Network Configuration -- WiFi discussion

I'm starting a new thread seeing as the last one took a different direction and never really got resolved.


As it stands, the current instructions for the installation guide make it challenging for users who need to connect to the internet via a Wireless connection rather than wired. The Installation_guide#Connect_to_the_Internet section essentially assumes the users are using a wired connection, and to test it with a ping. Otherwise, they are directed to the Network_configuration page. Unfortunately, that page does not provide clear instructions for new users (or even those who just want to do a quick connection) to establish a connection via wireless. Equally so, should users find their way to the Wireless_network_configuration page, there is some digging to do there in order to find instructions to setup a connection. These are currently using the iw command, which may prove to be challenging to some. I see a few different possible solutions to improve the user experience:

  • Add explicit instructions to the installation guide
    • This is not ideal, as it adds another place to maintain likely duplicate information
  • Add a reference to the Wireless_network_configuration page
    • This is better, though the page would likely need a new section, or some tweaking to allow users to more easily find the information they need to get setup
  • Add new/better instructions to the Network_configuration page
    • This may also prove to be tricky, seeing as that page is already fairly monolithic, and focuses mostly on wired connections

One other consideration (of which I also don't see any progress) is the discussions revolving around moving and breaking down the Network Configuration guides, to separate Wired and Wireless content. With this move I could see such instructions being provided there. In any case, the guide should provide instructions that are:

  • Easy to follow, particularly for new users
  • Puts no emphasis on persisting configurations, as this is not applicable during the install phase
  • Offers options (choice is King)

--CubeTheThird (talk) 23:47, 26 August 2018 (UTC)

The Wireless network configuration link was removed because it cannot be used independently of Network configuration. I propose the following:
  1. Implement Talk:Network configuration#Moving Ethernet-specific sections to Wired subpage. Network configuration currently isn't straightforward. The actual setup instructions are hidden in the Network management section and it's confusing that wireless has a subpage but wired and medium-agnostic configuration are mixed together. See my demo.
  2. Have the Connect to the Internet section only link Network configuration and move the dhcpcd udev rule note there.
  3. Move Wireless network configuration to Network configuration/Wireless and move its iw section to a dedicated article because since recently we also have iwd.
The result should be more user-friendly without duplicating content.
--Larivact (talk) 07:13, 27 August 2018 (UTC)
As for your demo, note that in Talk:Network_configuration#Ongoing_rewrite, Alad said: "ping is one of the very first commands a new user has to run on installation to verify the availability of an internet connection". So unless you intend to direct users from the installation guide directly to the Network configuration#Troubleshooting section, there is still some more thinking to be done... -- Lahwaacz (talk) 07:21, 28 August 2018 (UTC)
Well then let's keep Connect to the Internet and revise it:

The installation image has a udev rule that enables the dhcpcd service for Ethernet network interfaces on boot. If you use Ethernet, verify the connection with ping:

# ping archlinux.org

If the ping fails see Network configuration#Troubleshooting. If you want to use Wi-Fi or a static IP address, stop the dhcpcd service with systemctl stop dhcpcd@interface where the interface name can be tab-completed and proceed with Network configuration.

--Larivact (talk) 08:41, 28 August 2018 (UTC)
That looks good to me. -- Lahwaacz (talk) 14:57, 28 August 2018 (UTC)

Wording in example layout table and size of EFI partition

Wording in example layout table

Regarding Installation_guide#Example_layouts:

even if many users will understand remainder of the device as what is left after size of /dev/sdx1 and /dev/sdx3 are subtsructed from the size of the device, I think the order of the table might be confusing for some. Some people might set /dev/sdx2 to the size of the device minus size of /dev/sdx1, and then stumbled at where from 512 MiB, or larger, are to be found for /dev/sdx3. Either suggest the swap space as /dev/sdx2 and / as /dev/sdx3, or better explain the meaning of the remainder of the device for sdx2. Regid (talk) 14:05, 29 December 2018 (UTC)

Using /dev/sdx2 for swap is questionable, it doesn't emphasize that swap is optional nor is it consistent with other articles like dm-crypt/Encrypting an entire system. If you have some better explanation for "remainder of the device" feel free to propose it. -- Alad (talk) 08:55, 30 December 2018 (UTC)
I was trying to say that a user that patitions his HD by following the table might do the following: look at first row in the table, and creates the EFI partition. Than continue with the 2nd row. So he creates a partition at the remainder of his HD. Now he comes to the 3rd row: where will he get 512 MiB, or larger? As for dm-crypt/Encrypting an entire system, I might be wrong thinking that each partition is considered a separate device, so it doesn't matter if the swap space is before, or after, the / partition. Regid (talk) 12:41, 30 December 2018 (UTC)
Perhaps call it "Principal part of the device" instead. NB the term "remainder of the device" is often used for a separate /home in other articles. --Indigo (talk) 21:33, 3 February 2019 (UTC)
Partioning disks mentions fdisk but better is gdisk /dev/sdX (x representing your drive. mine is sda) then running commands x, z, y, y after that run cgdisk Drillsar (talk) 01:04, 16 May 2019 (UTC)

Confusing partition numbers

The partition numbers on this page are confusing. The table calls the root partition /dev/sdX2, while the text below calls it /dev/sdX1. It should be matched for consistency. Sabinscabin (talk) 20:29, 13 April 2019 (UTC)

Probably better to do it as [7] -- Alad (talk) 18:53, 1 May 2019 (UTC)

Incomplete listing of keymaps

When setting the keyboard layout, the command # ls /usr/share/kbd/keymaps/**/*.map.gz does not list all keymaps, as e.g. i386 has another subdirectory which breaks the regexp pattern. Adding another wildcard character # ls /usr/share/kbd/keymaps/**/*/*.map.gz shows the rest of the keymaps, back won't the keymaps from the former command. How about using ls -R /usr/share/kbd/keymaps/* | less -- Josephgbr (talk) 10:25, 31 July 2019 (UTC)

Are you sure ls /usr/share/kbd/keymaps/**/*/*.map.gz is correct? ** recursively matches subdirectories, so **/ and **/*/ should be the same.
$ ls -l /usr/share/kbd/keymaps/**/*/*.map.gz | wc -l
216
$ ls -l /usr/share/kbd/keymaps/**/*.map.gz | wc -l
216
Too bad that grml doesn't set GLOB_STAR_SHORT, the command could be simplified to ls /usr/share/kbd/keymaps/***.map.gz.
Personally I think, if we're making it more complex why not use basename -s .map.gz /usr/share/kbd/keymaps/**/*.map.gz, it would omit the paths and only output values that are usable by loadkeys and /etc/vconsole.conf.
-- nl6720 (talk) 15:33, 31 July 2019 (UTC)
Indeed, I probably skipped the second asterisk ** and didn't even notice. Your suggestion, although nicely remove path from the output, seems a little bit more complex and prone to typos. But if using it, it looks awesome prefixed with | column -- Josephgbr (talk) 16:04, 31 July 2019 (UTC)
I don't think it's that more complex, since the typed path remains the same. Updated [8], thanks. -- Alad (talk) 23:59, 5 September 2019 (UTC)
Reverted again since it's better to describe leaving out file path and extension, than not and hide it behind an additional step (basename). -- Alad (talk) 13:08, 7 September 2019 (UTC)

HiDPI on the console

With an ever increasing number of HiDPI displays, including at the begging of the article a section about adjusting the scaling factor or changing the font can be helpful, see HiDPI#Linux_console. Goetzc (talk) 02:21, 8 August 2019 (UTC)

It could be added as an example for setfont in Installation_guide#Set_the_keyboard_layout. The issue I have is that HiDPI#Linux_console mentions that tty2-6 may be unusable, while the Installation guide specifically instructs to change ttys as required in Installation_guide#Boot_the_live_environment. -- Alad (talk) 13:07, 7 September 2019 (UTC)

First mention of /mnt in example partition layout

/mnt is mentioned at mount point in Installation_guide#Partition_the_disks, while /mnt is made explicit two sections later in Installation_guide#Mount_the_file_systems. As I recall it, this was changed because some users blindly copy pasted commands and mounted /boot on the live system, instead of /mnt/boot. Some options:

  • Introduce another column describing the mount point on the installed system.
  • Actually explain /mnt early.
  • Revert the "mount point" to not include /mnt.

-- Alad (talk) 13:03, 7 September 2019 (UTC)

I don't understand what's the actual problem here... -- Lahwaacz (talk) 09:36, 8 September 2019 (UTC)
From what I read on #archlinux-wiki, this comes from https://www.reddit.com/r/archlinux/comments/d0v0j3/is_it_just_me_or_is_the_prospect_of_installing/ where the user was confused by the lack of root mountpoint (i.e. /mnt vs /). A question could be raised, if we should concern ourselves with users who have strong opinions about the wiki content yet can't be bothered to propose improvements in the talk pages...
About Alad's proposed options: I disagree with the first option, I think it will just complicate things even further. I support the third option and maybe adjusting the column header like in Special:Diff/581800.
I'd actually would like to go even further and change the commands run from outside chroot to be visually distinct, e.g.:
root@archiso # mount /dev/sdX1 /mnt
I think it would better solve the underlying issue.
-- nl6720 (talk) 15:26, 8 September 2019 (UTC)
I'm not overly fond of the longer column name. For the last proposed option, I may agree if this is formalized in Help:Style, so that it is not specfic to the Installation guide. -- Alad (talk) 11:20, 10 September 2019 (UTC)
Adding it Help:Style was my intention, since other articles, too, will need to use that style for some commands. I'm thinking of creating a template for it: Special:Permalink/581945. -- nl6720 (talk) 10:19, 11 September 2019 (UTC)
Sounds good to me, I'd just prefer the regular (non-bold) font for the prompt as above. -- Alad (talk) 21:54, 13 September 2019 (UTC)
Special:Permalink/582327. Are there any other opinions about creating such a template? Or should I take this discussion to Help talk:Template per Help:Template#Creation? -- nl6720 (talk) 18:31, 14 September 2019 (UTC)
  1. How are you going to call the template? This template would probably add to the Help:Template#Code formatting templates series, should it be named in a consistent fashion?
  2. Should this template support custom prompts, and if so, should it be called "pc" (from "(custom) prompted" code)?
  3. I don't like the red color too much, if bold is not an option maybe we can go green|purple|blue, something that recalls less a warning of some kind? Or can we just leave it with the default font color? Or a slightly fainter black?
  4. I haven't looked well into it, but maybe we can instead add an optional argument to Template:bc and Template:hc that prefixes a custom (colored) prompt? I wouldn't see a problem with repeating "root@archiso #" in every instance, or we may derive the new template from those two at that point.
  5. The template should probably be derived from Template:bc in any case, for simpler code, see Template:Sandbox2.
-- Kynikos (talk) 17:36, 16 September 2019 (UTC)
  1. Initially I was going to call it Template:Archiso since it would be Archiso-specific, but I'm starting to think that creating a more general-purpose template would be better. It could then be used in PostgreSQL and the [postgres]$ convention would get formalized in Help:Style. Now the issue is the [user@peer-a]# in Template:hc used in WireGuard. I'd rather not create two new templates, but I'm having trouble getting Template:Sandbox to work :(
  2. I like your "Template:pc" suggestion.
  3. Be glad I didn't post my first draft that was slightly more colorful. From your offered colors, I'd choose purple.
  4. I'd rather not mess with the established templates just for this change, so I'd prefer creating a new template.
  5. I didn't even think about using Template:bc. Is it a good idea to do that? The new template might need to be updated if Template:bc is ever changed in an incompatible way.
-- nl6720 (talk) 07:33, 17 September 2019 (UTC)
Yeah, after viewing your attempts and looking into it myself, I think modifying bc/hc is out of discussion, it would add too much code/style for so little use.
Thinking about this again one day after, I feel I'm realizing that my concerns in general may descend from the fact that we're going to create a template to represent (block) code, even though we already have 2 which basically do the same thing, including allowing to include a prompt; the only addition of this "Archiso" or "pc" template would be the formatting around the prompt, so why not keep it simple (I know, "simplicity" is often subjective and controversial) and instead either make a Template:Archiso to be used like {{bc|{{Archiso}} mount /dev/sdX1 /mnt}} or Template:ps (or Template:PS) to be used like {{hc|{{ps|root@archiso #}} mount /dev/sdX1 /mnt}}? They also work with Template:hc and space-prefixed code blocks!
Putting the choice of color aside, if the above idea of a standalone prompt template isn't welcome, I think my second choice would be to make two Template:pbc and Template:phc that work like {{pbc|$|ls}} and {{phc|$|ls|...}}, with the style rule to use them only in case of complex prompts. I'd still derive them from bc/hc to inherit any changes that we'd decide to make to them, and avoid repeating that ugly <pre> hack even more.
Otherwise I give up and accept the Template:Archiso that works like {{Archiso|mount /dev/sdX1 /mnt}}, in the hope that one day we won't need an analogous "hc" version.
-- Kynikos (talk) 14:24, 17 September 2019 (UTC)
I can't say I really like the idea of {{bc|{{Archiso}} mount /dev/sdX1 /mnt}} or {{hc|{{ps|root@archiso #}} mount /dev/sdX1 /mnt}}. I'd prefer creating Template:pbc and Template:phc.
I still don't get what's wrong with Template:Sandbox. It should just work:
prompt # command
code
-- nl6720 (talk) 04:43, 18 September 2019 (UTC)
FWIW (and a bit of fun) I've fixed Template:Sandbox, although I'm not sure if we really need that level of automation ^^ I stick to my position above, is there a third (or more) opinion? -- Kynikos (talk) 15:48, 18 September 2019 (UTC)
I think you like the #800080 shade of purple, right? ;-) Lahwaacz (talk) 11:39, 21 September 2019 (UTC)
Yes, I do like that one :D but I think it would be too bright for this template. -- nl6720 (talk) 11:52, 21 September 2019 (UTC)

Suggest minor change to the umount step

Admittedly quite a minor point, but the last step before Chroot is "Check the resulting file in /mnt/etc/fstab afterwards, and edit it in case of errors.".

As such, assuming the user does this, it is quite likely that they will be in /mnt/etc right before they Chroot, and will be returned to that directory when they get down to "Exit the chroot environment by typing exit or pressing Ctrl+d.".

The next (optional) step is to run umount -R /mnt, which will will fail unless the user moves out of /mnt (because the mount is still in use).

This happened to me when I was following these instructions, and (even being reasonably familiar with Linux) it threw me for a couple of seconds.

As I say, it's quite a minor point but, just in the interests of being as user friendly as possible, perhaps a change..?

Suggest changing it to "Optionally change back to the root of the live system with "cd /", and then manually unmount all the partitions with umount -R /mnt: this allows noticing any "busy" partitions, and finding the cause with fuser.

—This unsigned comment is by Lewis2e (talk) 14:51, 2 October 2019 (UTC). Please sign your posts with ~~~~!

You don't need to cd to /mnt/etc/ just to be able to edit /mnt/etc/fstab. -- nl6720 (talk) 14:59, 2 October 2019 (UTC)
I didn't say that you did... but OK. I withdraw my comment. Lewis2e (talk) 17:32, 2 October 2019 (UTC)

Changes for the base package

Installations without base

The base group was replaced with the base package: [9] This change was reflected in Installation_guide#Install_the_base_packages with [10]

With [11], I removed the sentence "We only officially support installations that have the base package installed." because it opens a new rabbit-hole when something is "officially supported" in the installation guide, is not. With this sentence included, pretty much anything (including "installations" that are not, or only partially, followed from the Installation guide) may be supported merely from having the base package installed.

On the other hand, some notion that removing the base package results in an installation that is "not Arch" makes sense, but we should discuss on the best approach on doing this. -- Alad (talk) 10:22, 6 October 2019 (UTC)

pacstrap suggested packages

linux-firmware has been made optional earlier this week in the linux kernel, this together with the base package changes means that linux-firmware is not installed with the currently recommended pacstrap command. I would argue that most people expect linux-firmware (or need it, even).

I propose adding linux-firmware to the pacstrap command, with a note that this is only required on baremetal systems (i think?) -- Coderobe (talk) 10:46, 6 October 2019 (UTC)

Adding linux-firmware to the pacstrap command would mark it as explicitly installed, while it is just an optional dependency of the kernel packages. -- nl6720 (talk) 13:17, 6 October 2019 (UTC)
If the purpose of installing linux-firmware is clearly explained in the guide, what's the issue in having it explicitly installed? As I see it, if it's a bare metal installation, then it is something that should be considered as explicitly installed. -- CodingKoopa (talk) 16:15, 6 October 2019 (UTC)
Previously, all packages from the base group were explicitly installed... -- Lahwaacz (talk) 17:43, 6 October 2019 (UTC)
Yes that's true, but the base and linux changes now allow for proper dependency tracking, unfortunately there's no way to tell pacstrap that a package should be installed as a dependency. Anyway, forget about this complaint of mine, there are more pressing issues. -- nl6720 (talk) 05:36, 7 October 2019 (UTC)
New Arch user here. I don't really know what a "baremetal system" is (1), but after rebooting, I ended with a system that seemed to be "frozen" in the middle of the boot process. Actually it was not, just without working graphical drivers. I've got some experience in Linux and, after checking journald logs from within a chroot and remembering I watched a similar problem in a Debian Testing installation some months ago, I knew it was due to missing linux-firmware. So, I guess this is an issue another users are going to suffer and probably the less experienced won't be able to solve it.
(1) This is an old but reliable machine that I want to use as a backup system: a Core2 Duo with an AMD Caicos as graphics card.
—This unsigned comment is by Angrytomato (talk) 2019-10-09. Please sign your posts with ~~~~!
Thanks for the feedback, I think we need to include linux-firmware next to the kernel package. I've made a quick draft which also more clearly lists other packages that may be necessary. -- Lahwaacz (talk) 21:33, 9 October 2019 (UTC)
linux-firmware can also useful in virtual machines when passing through a piece of hardware that requires it (GPUs being the most common case). -- nl6720 (talk) 04:20, 10 October 2019 (UTC)

Oh, addendum: base does *not* contain a text editor. This should probably be made clear in a similar manner as linux-firmware. New users will find themselves without a regular text editor when booting into the installed system otherwise. -- Coderobe (talk) 11:54, 6 October 2019 (UTC)

The base package also doesn't depend on any file system userspace utilities. The installed system would need them for fsck. -- nl6720 (talk) 13:15, 6 October 2019 (UTC)
I suggest also letting users know about manual pages and recommend adding: man-db man-pages texinfo.Tinywrkb (talk) 15:29, 6 October 2019 (UTC)

All packages that are in base (group), but not in base (package):

$ comm -23 <(pacman -Sgq base | sort) <(expac -S "%D" base | tr -s " " "\n" | sort)
cryptsetup -- indirect dependency of base
device-mapper -- indirect dependency of base
dhcpcd
diffutils
e2fsprogs -- indirect dependency of base
inetutils
jfsutils
less -- indirect dependency of base
linux
linux-firmware
logrotate
lvm2
man-db
man-pages
mdadm
nano
netctl
perl
reiserfsprogs
s-nail
sysfsutils
texinfo
usbutils
vi
which
xfsprogs

Note man-pages.

-- Lahwaacz (talk) 17:11, 6 October 2019 (UTC)

e2fsprogs is not in base, but it gets installed as a dependency. systemd -> audit -> krb5 -> e2fsprogs. Grazzolini (talk) 13:15, 7 October 2019 (UTC)
So what do we do? It feels like we're making policy decisions, which shouldn't really be happening. -- nl6720 (talk) 06:00, 7 October 2019 (UTC)
We might need an additional section like "Install additional packages", either in Installation guide#Installation or Installation guide#Configure the system or somewhere in General recommendations. -- Lahwaacz (talk) 11:21, 7 October 2019 (UTC)
General recommendations would be too late for networking packages, but man-db, man-pages, texinfo, diffutils and less could be mentioned there.
Networking package installation would belong nicely in Installation guide#Network configuration and there already is a precedent for placing package installation in Installation guide#Configure the system with Installation guide#Boot loader (although that section is regularly criticized).
File system userspace utilities: Installation guide#Fstab is before chroot, so not there. Or maybe split Installation guide#Fstab into two parts (in archiso and in chroot), but that would needlessly complicate things. Maybe hint about it in Installation guide#Format the partitions and place the installation instructions in Installation guide#Installation?
Now, text editors ... After Installation guide#Chroot a text editor is needed otherwise users have to create and modify files with echo and sed which, although doable, is ridiculous. Maybe place it in Installation guide#Installation? While I feel that recommending any specific text editor would just lead to bikeshedding or perpetual complaints, but IMHO not recommending one would again be needless complication.
-- nl6720 (talk) 12:23, 7 October 2019 (UTC)
Would be nice to get a draft in this time, before merging the changes to the article (mea culpa on allowing editing of the installation guide without prior consultation.) -- Alad (talk) 15:17, 7 October 2019 (UTC)
I don't think any of my changes besides Special:Diff/584615 were that controversial. There's no simple way create a draft unless there's a consensus on what to put where. Otherwise you'll just get a draft for the whole page, which I consider to be a waste of time. -- nl6720 (talk) 15:33, 7 October 2019 (UTC)
I meant the things you mentioned in your last reply (text editors and all that stuff). Doesn't have to the whole page, just specific sections. -- Alad (talk) 08:02, 8 October 2019 (UTC)
I've made a quick draft, if you think it's an improvement we can merge it and then discuss if individual recommendations should be moved elsewhere. -- Lahwaacz (talk) 21:33, 9 October 2019 (UTC)
I'd like to see the note changed to a tip, but otherwise I support your draft. Another thing to consider, in Special:Diff/584590 I changed "specific wireless firmware" (i.e. Broadcom wireless) to "firmware files for devices". Maybe #Draft: Install essential packages should mention "specific wireless firmware". -- nl6720 (talk) 04:20, 10 October 2019 (UTC)
I've changed it, feel free to edit it further. -- Lahwaacz (talk) 06:46, 10 October 2019 (UTC)
Current version looks OK to me. -- nl6720 (talk) 08:04, 10 October 2019 (UTC)

Complicate Install for the Majority to Cater to the needs of a Minority

These seem like a lot of unnecessary changes to what has been a reliable install 'base' for the past decade. Better to leave the base group as it was and create a 'base-container' group for those that do not want a complete base install. Installing the new 'base' will force new installs to install 'base' + a list of multiple packages just to arrive at a running install. It just seems backwards to complicate the install for a majority of users simply to cater to a minority of container installs. David C. Rankin, J.D.,P.E. -- Rankin Law Firm, PLLC (talk) 18:35, 8 October 2019 (UTC)

Regardless of whether people agree or disagree with you, this is not the place to discuss whether "base" should be a package or a group and which packages it should pull in. The installation guide merely documents the reality, it does not enforce how Arch works. If you want to change how Arch works, please talk to the devs and submit a bug report, making sure that you understand the reasoning behind the change.
I'll close this to prevent bikeshedding on this talk page.
-- Lahwaacz (talk) 19:13, 8 October 2019 (UTC)

Draft: Install essential packages

Use the pacstrap script to install the base package, Linux kernel and firmware for common hardware:

# pacstrap /mnt base linux linux-firmware
Tip: You can substitute linux for a kernel package of your choice. You can omit the installation of the kernel or the firmware package if you know what you are doing.

The base package does not include all tools from the live installation, so installing other packages may be necessary for a fully functional base system. In particular, consider installing:

To install other packages or package groups, append the names to the pacstrap command above (space separated) or use pacman after the chroot step. For comparison, packages available in the live system can be found in packages.x86_64.

A bit long, but LGTM -- Alad (talk) 20:43, 10 October 2019 (UTC)
OK, merged. -- Lahwaacz (talk) 13:32, 12 October 2019 (UTC)