Fedora tested in c't: grub missing feature

Jeff Spaleta jspaleta at gmail.com
Thu Feb 5 20:02:22 UTC 2009


2009/2/5 Thomas Bendler <ml at bendler-net.de>:
> ??? You are not able to figure out what a root partition is and what a boot
> partition is? The only thing you need is the root partition, the rest will
> be in /etc/fstab. And to find out what a root partition is, is not really
> complicated, all distributions will have a common set of directories in /,
> if not, they are not detected, but for most of the distributions this will
> work.

I know a lot of things.. that the software in my computer does not
know. I'm smarter than my computer. You are smarter than your
computer.   What I know about the state of my filesystem is in fact
partly conceptualized in my human experience and not recoverable from
a blank slate reading of the contents of the filesystem.

We are talking about teaching anaconda to know what I know in a
consistent and fast algorithmic way. if it takes more time for
anaconda to do it, by bruteforce disk searching of the filesystem
layout, than it takes for someone to edit the grub config by
hand...what the hell is the point. There's no win there.

If we are going to find an implementation that works then we need to
think about it like a computer. We need to be dumb as dirt
(technically sand as we still have silicon based hardware), we need to
require explicit instructions...instead of making intuitive fuzzy
leaps of logic that involve heuristic parsing that relies on apriori
situational knowledge.
I hand you a harddrive, I tell you it has 15 partitions sitting on the
disk, one of them is an ntfs partition and its the only partition
flagged as bootable.  List the set of explicit instructions that
detail how you would tell me find which of those partitions represent
a distinct linux operating system and how to generate a working grub
entry for each one of them.


-jef




More information about the devel mailing list