2009/12/11 Per Jessen
Rob OpenSuSE wrote:
Are there any good reasons why there cannot be supported boot configurations :
1) Normal partitions or RAID1, / ext3/ext4 with /boot sub-directory
Shouldn't be a problem.
2) Normal partitons or RAID1, / "exotic" file system, with /boot small ext2/ext3/ext4 partition
Ditto.
3) ReiserFS, currently somewhat broken; fix desireable 4) LVM, currently "exotic"; direct support desirable 5) btrfs, currently "exotic"; direct support desirable
I _think_ they all work with lilo, please correct me.
lilo's weakness becomes a strength in these cases due to simplicity of blocklists. But remembering how hard it was to support ppl on mail list & forum in past, I'm afraid lilo(8) just is going to have significant support costs for the masses. When GRUB was introduced, the typical users really did like it and prefer it.
With Installer warning when a possibly broken boot partition is chosen by the expert user.
Hmm, do we need to warn the _expert_ user?
It would seem so, from reading this Mailing list. Ways to get totally reliable configurations are ignored, and this results in trips to Bugzilla. If you don't tell the end-user they are doing something unreliable during an installation (like is done if you pre-mkfs system partitions and do not allow format); they very likely in their expertise expect their flakey configuration to simply work and get upset when it doesn't.
lilo - http://freshmeat.net/projects/lilo/ * static - uses map files * confused end users even when it was most popular boot method - poor diagnostics on failure - poor recovery on failure - less friendly config file than GRUB 0.97
Rob, your bias is showing :-). Let me add some positive points:
* just works, hardly ever fails. * easy text-file config * tried and tested.
No, lilo(8) never "just worked" for people with less stable systems. Remembering the effort needed to explain things like booting windows with lilo, using Bios Disk codes for example, I do think it is fair to say GRUB is easier. There were good reasons, why the distro moved to GRUB "legacy", and it was a popular decision, and regarded as a great improvement. Inflicting lilo(8) on end users again, will cause a "revolt". I actually liked lilo(8) and regarded myself as a relative expert on it, I was dubious of GRUB's requirement for boot time filesystem support, but it did not take long for GRUB's advantages to become apparent. I also do not miss explaining lilo's error codes to ppl, or why they cannot boot, because they forgot to make map files. There really is a reason, why the BSD's and Solaris chose GRUB over their own variant of lilo. lilo IMO makes a good fall back boot loader, but when things go wrong because of user errors, or hardware errors the diagnostics are awful, and that increases support costs. As lilo's last release is 2007, I would have doubts about support for the newer boot methods like EFI http://en.wikipedia.org/wiki/Extensible_Firmware_Interface Also according to the links I referenced, lilo's architecture support is narrow, which might not seem important today, but could be if say ARM netbooks take off in next few years. Rob -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org