On Mon, 5 Dec 2011, Julian Elischer wrote: > On 12/4/11 9:21 PM, Daniel Eischen wrote: >> On Dec 4, 2011, at 7:42 PM, Julian Elischer<julian_at_freebsd.org> wrote: >> >>> On 12/4/11 3:36 PM, Randy Bush wrote: >>>>> This seems too reasonable a suggestion, but, as always, the devil >>>>> is in the details. There will be long. painful discussions (and >>>>> arguments) about what to remove from the base to the new structure >>>>> and what things currently NOT in the base should be promoted. >>>> as one with a long list of WITHOUT_foo=YES in /etc/src.conf, this is >>>> tempting. but, as you hint, is this not just doubling the number of >>>> borders over which we can argue? >>>> >>>> but let's get concrete here. >>>> >>>> i suspect that my install pattern is similar to others >>>> o custom install so i can split filesystems the way i prefer, >>>> enabling net& ssh >>>> o pkg_add -r { bash, rsync, emacs-nox11 } (it's not a computer >>>> if it does not have emacs) >>>> o hack /etc/ssh/sshd_conf to allow root with password >>>> o rsync over ~root >>>> o hack /etc/ssh/sshd_conf to allow root only without-password >>>> o rsync over my standard /etc/foo (incl make.conf and src.conf) >>>> and other gunk >>>> o csup releng_X kernel, world, doc, ports >>>> o build and install kernel and world >>>> >>>> and then do whatever is special for this particular system. >>>> >>>> anything which would lessen/simplify the above would be much >>>> appreciated. anything not totally obiously wonderful which would >>>> increase/complicate the above would not be appreciated. >>> my suggestion is that the 'sysports' or 'foundation ports' or >>> 'basic ports', (or whatever you want to call them) in their package >>> form come with the standard install in fact I'd suggest that they >>> get installed into some directory by default so that 'enabling' them >>> ata later time doesn't even have to fetch them to do the pkg_add. >>> >>> They have pre-installed entries in /etc/defaults/rc.conf. and only their >>> rc,d >>> files need to beinstalled into /etc along with their program files. >>> They are as close to being as they are now with the exception of >>> being installed in the final step instead of at the same time as the rest >>> of the stuff, >>> and it allows them to easily be 'deinstalled' and replaced by newer >>> versions. >> I really don't understand how this is much different than having them exist >> in base. We have WITHOU_foo (I don't really care if that were to become >> WITH_foo if we want to default to a more minimum system), so one can always >> use ports if they want some different version of foo. And it's not just >> releases we care about, we want a stable foo (BIND for example) with >> security and bug fixes throughout all updates to -stable, not just at >> releases. >> >> I want to do one buildworld and have a complete and integrated system. I >> don't see how having a separate repo for sysports helps; it is yet another >> thing I have to track. And are ports in sysports going to default to being >> installed in / or /usr/local? > > I think there are several differences.. > 1/ The ability to UNINSTALL it and replace it completely with a differnet > version If we go to a complete pkg-based system, then there is no difference here, so why not do that? > 2/ allow easy leave-out feature.. leaving it out is less risky.. WITH_FOO/WITHOUT_FOO vs pkg_delete, not sure there is much of a difference. The advantage of WITH/WITHOUT is that the system is built as a whole and integrated. src/ developers are suppose to not break src/; they may not be so inclined to worry about sysports. Will emphasis be put on src/ developers to include sysports in their "buildworld" and will tinderboxes also include sysports? > 3/ probably the most important.. allowing both ports and src developers to > work on the packages. Give ports maintainers that maintain BIND, FOO, access to src/ (which they probably have already). > 4/ allowing us to promote some of the commonly used packages to a more > supported level without actually bringing them into the base system. -- DEReceived on Mon Dec 05 2011 - 12:36:40 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:21 UTC