XPost: linux.debian.maint.boot
Cyril Brulebois <
kibi@debian.org> writes:
Hi,
ValdikSS <iam@valdikss.org.ru> (2021-04-28):
eatmydata-udeb package is designed to speed up the installation process.
However, it's not used by default and could be activated only with preseed >> file or kernel cmdline argument.
Please consider increasing eatmydata-udeb priority to standard in
debian-installer overrides, for it to be used by default. This will speed up >> installation by an order on slower HDDs.
It will be perfect to include this change before final Bullseye ISO release. >>
For more information, read the post in debian-boot:
https://lists.debian.org/debian-boot/2021/03/msg00121.html
We already have a lot of serious problems to solve in the installer;
I wouldn't want to see possible extra perturbations due to a syscall interceptor.
The best would be to have that kind of change happen right at the
beginning of a release cycle, rather than in the last few weeks
before a release…
A compromise might be thinking about backporting the change to
bullseye once it's been tested with some D-I Bookworm Alpha 1 (but
that would also be a rather important change to backport to a stable release…).
Happy to hear other opinions.
Would it be possible to make this conditional on something being set on
the kernel command-line, so that people are in a position to opt-in to
using it?
This seems to be an endlessly repeating dance, where the effort
available to d-i seems to ebb and flow in synchronisation with the
release cycle, but the opportunity to make interesting changes is in
anti-phase with that.
If we had a conditional looking out for e.g. "experimental=..." on the
kernel command line, and then treating that as a list of things that
people are trying to test, then people would be able to easily
experiment with these things and report successes/failures such that we
might be able to:
take advantage of people that are keen to make contributions during the
late phase of the release that can spend one cycle in a state where
they are only used if the user opts-in.
provide a trivial way of taking advantage of innovations without
endangering the reliability of the default installer, where those
supporting the change just need to add 'experimental=eatmydata' to the
kernel command line in order to give confidence for inclusion in the
next release.
Obviously, there is a problem introducing such a change right now.
Typical, eh ;-)
Could we start out with a baby step having a new experimental-netinst
flavour of ISO, which includes a list of experimental udebs in addition
to the normal ones?
I'm happy to put effort into making that happen BTW.
Cheers, Phil.
--
|)| Philip Hands [+44 (0)20 8530 9560] HANDS.COM Ltd.
|-|
http://www.hands.com/ http://ftp.uk.debian.org/
|(| Hugo-Klemm-Strasse 34, 21075 Hamburg, GERMANY
--=-=-Content-Type: application/pgp-signature; name="signature.asc"
-----BEGIN PGP SIGNATURE-----
iQIzBAEBCgAdFiEE3/FBWs4yJ/zyBwfW0EujoAEl1cAFAmCJJSsACgkQ0EujoAEl 1cC3IRAAn79jBoUlc4X0jpmPmf0xDbYCCOSr811f/X9zmQwb9Et1BKmj5yLAjOD1 OI40tJ6r7imvyDgzwh4loZxWba6VGAhRKebbOOjVedIvC7N3glAUJ4IdprzxaK0b bdCs8FApj8KgyQmyKO1fxgzEcWZhu8pApVm8j9d4n+vzIsPl98FFdxd+HlXAXRQ7 j6TFRIi9elhEPtS9yvFdKkp1VZBuoylC5AqCTyUMEbe/Q6AuRyYYf/ohj+ctJk+8 Xm5B/XEAuB1zn2XdN2aDTbsglgrNeai4g94YQEvxJKaLORYXA0ez+WM8ncdZWC2+ Zd3Dxp+QM7yN8cHeVhc3ReybHSk+kJ389bdyivRga4I3jH4n3xr4/ZBhUXzZ9s24 CtTJn7FpriuExmKj9p19HDsCaGRjNOGAOdgWBHSumOhiSc4euOB6Oq8aGB3BsR1m fG4QoOuxKNcNkvI5JPMn+ytlsy7+3jZfbxBX7uYeohBTXsQrKSjp58M61QgBAedB wV2VD7fRE249VJV3u5yuKu74KqCWHybj4TM6SDaW0pP4yRpGNfIvQo+xNxgG3V0M bohJygtt1Wl/puJ/w4JIhpcohLmtjBxevfIXAo9KRYohScZmG4hslEow+w37jgQf /DLE4oLoOW8Nujt