[syslinux] syslinux vs grub
H. Peter Anvin
hpa at zytor.com
Sat Jan 8 18:33:58 PST 2005
Patrick J. LoPresti wrote:
>>What I really mean is I'm still debating whether or not to develop
>>syslinux further, whether or not I should develop it further but in
>>a different direction, or take the project to maintenance-only mode.
> You asked what it would take to replace GRUB. For that, I do think
> handling file systems other than FAT and ext2 is a requirement. Last
> I checked, SUSE still defaulted to reiserfs. And I have seriously
> considered moving my own systems to XFS. If you do not want to code
> support for each file system yourself (and I don't blame you), then I
> would suggest providing some kind of framework where other motivated
> hackers can add support for their favorite file systems.
> File system "modules" was just an idea, as was the command-line stuff.
> They would be useful for floppy boot disks, which I myself have found
> handy once or twice. But I could still use GRUB for that if my system
> were normally booted by EXTLINUX (or XFSLINUX or whatever).
More filesystems isn't a problem; the framework for that is already
there, and with the EXTLINUX work I did a lot of (necessary) cleanup
that allows disk-based filesystems to share code. So supporting XFS and
Reiser is probably not that hard, although I need to make symlinks work
in EXTLINUX first.
It's supporting multiple filesystems *at the same time* that really and
thoroughly breaks the syslinux model. This could mean it's the wrong
model, and that's what I'm trying to assess.
>>In particular, I'm starting to seriously questioning the value of
>>some of the advanced features I have on the drawing board (and in
>>some cases, partially implemented) when built on top of the syslinux
>>framework. I don't want to spend a lot of time building those new
>>features if they are ultimately useless.
> Well, I guess it depends on what your goals are. GRUB is certainly
> ripe for replacement in a lot of ways, IMO. It is bloated, the
> maintainers are largely non-responsive, and the whole project is
> pretty much stalled. If you want to try to supplant it, I think you
> will find lots of people rooting for you and wanting to help where we
Part of what I'm trying to do is to figure out what the goals should be.
Building on the syslinux platform, I think I can do a lot of things
*except* support multiple filesystems.
For one thing, I'm well on the way to have a graphics subsystem for
com32. It doesn't take that long to port standard libraries to libcom32.
More information about the Syslinux