[PC-BSD Testing] Rolling release criticism
jpm820 at gmail.com
Tue Dec 17 07:10:54 PST 2013
There are some great points here. I remember with the old pkg_add that
for a release by default FreeBSD would only fetch release packages. I'm
not sure if that is the case know as I think they have weekly updates
with the FreeBSD pkg repo now?
However maybe it would be wise to have build one big RELEASE packageset
at the time of RELEASE and never update it other than critical fixes.
Instead have the user opt in by using the chbranch function to change
the the STABLE release if they want bi-monthly package updates along
with the normal montly world and kernel updates. I was thinking maybe
this approach might free up some resources by not having to stay on top
of two packagesets as much.
I also agree FreeBSD needs more people running HEAD/CURRENT. This
benifits the project more than people using the old stuff but it also
turns each user into a beta tester potentially against their will. By
forcing users to switch to at least the STABLE branch to get new
packages it would force them to test the montly updates to world and
kernel as well.
On Tue, 2013-12-17 at 08:12 -0500, Claudio L. wrote:
> On 12/17/2013 06:13, Sam Fourman Jr. wrote:
> > The FreeBSD ecosystem needs MORE people running HEAD/CURRENT as well
> > as the VERY top of the ports tree, this is how you attract developers
> > and power users... and these people are the ones that will help
> > improve FreeBSD/PC-BSD for the general population.
> Precisely, but the only way you get more people to use it, is if it
> actually works. What I mean is the basic system should always work. Then
> you can test individual features that you **want** to install and test,
> but you probably shouldn't be testing all at the same time and without
> your consent:
> * The package manager
> * Hardware drivers
> * Basic web browsers
> * Basic services like Virtualbox
> These are basic things that should always work fine, then I agree with
> the testing of individual features.
> Imagine you are trying to test a new feature in 10.0, but you can't
> because you are wasting time fixing your browser, your packages, etc.
> You should have everything else working and out of your way so you can
> focus on testing PCI-PASS THROUGH on bhyve, not more, not less.
> If things break too much, you'll alienate users and they will leave to
> other more stable OSes.
> For example, we could have other packages updated fast, but the package
> manager only changed once or twice a year, for example. Just to make
> sure it works, and when the update gets pushed, it HAS to work for
> Testing mailing list
> Testing at lists.pcbsd.org
More information about the Testing