Tails needs to use standard booting methods

Posted May 4, 2014 11:04 UTC (Sun) by gavron (guest, #96951)
Parent article: Tails 1.0 released

Tails would best serve its community of users by allowing flexible booting from USB sticks.

Sadly that is the one area its bootup is MOST lacking. See e.g. https://tails.boum.org/support/known_issues/index.en.html...

If you've seen Ubuntu's live-boot, debian's live boot, Fedora's live boot (stop me if I'm boring you) you'll wonder WHY you can't do the very same thing with Tails. It's because they insist on their own boot method, so you can image an ENTIRE USB STICK for tails and then it will only fail on the 20+ machine types listed above... or put it on a CD. A CD... in 2014.
(It also doesn't boot on QEMU...)

Tails should get someone who knows the boot process to fix their startup scripts so that they can play nice with a Grub2 ISO loop bootloader, and instantly work on a lot more machines.

So... good on them for all they do... and for making it to 1.00, but the failure to use commonly accepted tools for boot (grub) and allow commonly accepted methods (USB live) means for now they are a project in search of someone to understand that to get users to use you you must first boot on the users' existing systems using modern methods.

Cheers

Tails needs to use standard booting methods

Posted May 8, 2014 11:04 UTC (Thu) by intrigeri (subscriber, #82634) [Link]

First, we don't want to encourage users to put other live systems on the same stick they are using for Tails, as it basically boils down to trusting these entire systems not to corrupt Tails, which does not match our threat model. So, officially supporting booting Tails with the GRUB2 ISO loopback support is not something that would match our mission, nor be a service to our users. Also, we are not primarily targetting power users who are able to use configure GRUB by hand. And those power users who really want to do it, and understand the drawbacks, can actually do it: last time I checked, it did work.

Second, Tails does use Debian's live-boot, and we have contributed to it quite a lot in the past. So, it is unclear what you mean our "own boot method". We use live-boot + syslinux, like many (if not most) other GNU/Linux live systems. Nothing custom in there.

What is somewhat custom, and you might be refering to it, is that a boot medium created by Tails Installer has a GPT, instead of a legacy MBR. This design decision was made 1. to make it easier to later support UEFI, which is now coming close (see e.g. https://tails.boum.org/news/test_UEFI/ and the results on https://tails.boum.org/blueprint/UEFI/syslinux/); and 2. so that we have ways to detect if a Tails persistent volume is present, based on partition labels (MBR partition tables have no partition labels, and filesystem labels are not usable here since the persistent volume is encrypted), and provide a better experience to the user based on that. A problem with GPT is that some flawed firmware are not able to boot in legacy (non-UEFI) mode from a GPT device, which explains most of the boot failures documented on our known issues page. Most of these should be fixed once the UEFI support is released (most likely in Tails 1.1, June 10). I believe this will address your "must first boot on the users' existing systems using modern methods" point. Some other firmware fail to boot from a stick created with isohybrid + cat, which 1. is no surprise given how hacking this method is; and 2. affects all live systems that document/support this way of installing, nothing specific to Tails here.

Regarding "a CD": you probably mean a DVD, as Tails does not fit on a CD for a while. Many GNU/Linux distributions provide ISO image, and document "burning a CD" as the primary way to get their installation process started, so I'm not sure what the "in 2014..." unclear sarcasm really means in this context.

Regarding "doesn't boot in QEMU": many users boot Tails in QEMU, and I have never heard of any problems with it. Please send us a proper bug report (https://tails.boum.org/doc/first_steps/bug_reporting/), which merely stating "it doesn't boot" on LWN definitely is not. Thanks in advance!

To end with, I'm glad to inform your that many thousands of people successfully boot Tails every day, so surely the big picture is not as bad as you are drawing it. Still, the situation is certainly not perfect, and we would welcome input and feedback
about specific situations that Tails does not support well. We'll definitely consider it, as long as it fits into our design goals and threat model :)