[blfs-dev] A proposal
rob at landley.net
Mon May 13 22:39:56 PDT 2013
On 05/12/2013 02:35:59 PM, Bruce Dubbs wrote:
> Armin K. wrote:
> > The GNOME/GTK+ Project packages are becoming outdated and newer
> > package that are adjusted for them seem to be more or less "failing"
> > with older ones (See several NetworkManager threads on
> blfs-support). On
> > the other side, I use Systemd and my system isn't anymore compatible
> > with any of the desktops (I can only check if they work with
> Systemd not
> > with ConsoleKit).
> > I want to propose a systemd branch for BLFS, where we (for now, I)
> > upgrade all packages that now depend on Systemd and make existing
> > work with it, too.
> I'd like to make an alternative proposal. How about removing those
> packages from BLFS (e.g. Gnome) that need systemd and putting them
> a separate document that references the main BLFS packages as needed.
> There are many packages in BLFS that do not need or use systemd, but
> be used in a systemd environment. Doing the work of maintaining
> packages in two documents seems counterproductive to me. Links
> separate packages are not unreasonable. There are several links in
> to BLFS.
I suggested some time ago that the Gnome/KDE stuff be split out from
the rest of BLFS, because it was sucking up all the developers time but
not of interest to people who want to use xfce or fluxbox or similar.
(Or just make an elaborate non-gui dedicated system to be a mail server
I'd link to the actual post under
http://linuxfromscratch.org/mailman/listinfo/blfs-dev with my
suggestions of obvious cleavage lines to break BLFS into separate
books, but the mailing list archive is 404.
> The reason I do not want to use systemd in LFS is because it makes the
> boot process opaque. I think presenting what happens during boot is
> important educational goal of LFS. Being able to read short bash
> scripts and using separate programs like syslogd accomplishes that
> much more simply and visibly than systemd. There are also many parts
> systemd that are simply not required in a linux system, but are there
> support only large multi-user systems (e.g. cgroups). A basic system
> just does not need resource limiting, prioritization, accounting, and
> controlling of groups.
There's also bound to be a better way to do it than Pointy Hair Linux's
pet project. Long ago I decided I didn't like udev, and did mdev
instead. These days, devtmpfs does the most interesting bits of what
mdev did anyway
If there's a real need for a "oh please not systemd" project that stubs
out these strange dependencies or something, I can add it to my todo
list. (I'd rather somebody else did, of course...)
More information about the blfs-dev