Right now the workflow, from what I see, is to carve out a little space on the disk, extract ISO there and point the bootloader to boot from it. While it is a good idea and will work on most of the systems, it will be a challenge on stuffed machines. Or encrypded ones - it is not always easy to resize encripted partition.
Additionaly, the ISO is already there, on the machine, when you start (even if just downloaded), and the perfectionist in me can't stand such a waste.
What if we point to the ISO directly?
Yes, we'll have to deal with NTFS drivers, yes, we'll have to deal with unlocking BitLocker, surprisingly no, we won't have to deal with Secure Boot if we're smart, but yes, this will be a chunk of work.
What I'm prroposing is to dynamically compile something similar to Ventoy (form precompiled parts, obviously), but targeted to a single system and a single ISO; and put that as an .efi to be rebooted into.
Then, this .efi bootstraps the actual ISO, dealing with necessary drivers and encryption (remember, Linux can be encrypted too!).
And we can even sign this .efi, so we don't have to deal with Secure Boot.
Maybe we could do even less than Ventoy, as we don't need to tug the legacy of GRUB for a single ISO boot with ocasional password prompt.
That could also help with generalization. I haven't looked into the reason why some distributions fail to boot and why it is preferred for them to be specificaly added by you, so I might be missing something, but chaniloading should help with this.
What do you think? I just got this idea and have no experience with UEFI - NTFS - custom bootloaders to predict the limitations, but it sounds sane and all necessary components seem to work.
Right now the workflow, from what I see, is to carve out a little space on the disk, extract ISO there and point the bootloader to boot from it. While it is a good idea and will work on most of the systems, it will be a challenge on stuffed machines. Or encrypded ones - it is not always easy to resize encripted partition.
Additionaly, the ISO is already there, on the machine, when you start (even if just downloaded), and the perfectionist in me can't stand such a waste.
What if we point to the ISO directly?
Yes, we'll have to deal with NTFS drivers, yes, we'll have to deal with unlocking BitLocker, surprisingly no, we won't have to deal with Secure Boot if we're smart, but yes, this will be a chunk of work.
What I'm prroposing is to dynamically compile something similar to Ventoy (form precompiled parts, obviously), but targeted to a single system and a single ISO; and put that as an .efi to be rebooted into.
Then, this .efi bootstraps the actual ISO, dealing with necessary drivers and encryption (remember, Linux can be encrypted too!).
And we can even sign this .efi, so we don't have to deal with Secure Boot.
Maybe we could do even less than Ventoy, as we don't need to tug the legacy of GRUB for a single ISO boot with ocasional password prompt.
That could also help with generalization. I haven't looked into the reason why some distributions fail to boot and why it is preferred for them to be specificaly added by you, so I might be missing something, but chaniloading should help with this.
What do you think? I just got this idea and have no experience with UEFI - NTFS - custom bootloaders to predict the limitations, but it sounds sane and all necessary components seem to work.