announcement --- planned Yum replacement now ready for user testing

Ralf Corsepius rc040203 at freenet.de
Thu Jan 2 07:35:53 UTC 2014


On 01/01/2014 11:13 PM, Lars E. Pettersson wrote:
> On 12/20/2013 12:33 PM, Ales Kozumplik wrote:
>> On behalf of the DNF team I'd like to invite all the interested Fedora
>> users in trying out and testing DNF in Fedora 20. DNF is a tool that
>> aims to fully replace Yum by Fedora 22. Please check out the blog post
>> for more information:
>
> A question, I found the following on
> <http://akozumpl.github.io/dnf/cli_vs_yum.html>
>
> "dnf erase kernel deletes all packages called kernel
>
> In Yum, the running kernel is spared. There is no reason to keep this in
> DNF, the user can always specify concrete versions on the command line,
> e.g.:
>
> dnf erase kernel-3.9.4"
>
> So if I issue 'dnf erase kernel' all kernels will be removed, and I have
> no kernel anymore?
Apparently, this is what dnf does:

# dnf remove kernel
Resolving dependencies
--> Starting dependency resolution
--> Finding unneeded leftover dependencies
...
---> Package kernel.x86_64 3.11.10-301.fc20 will be erased
---> Package kernel.x86_64 3.12.5-302.fc20 will be erased
---> Package kmod-nvidia.x86_64 1:331.20-10.fc20.1 will be erased
---> Package kmod-nvidia-3.11.10-301.fc20.x86_64.x86_64 1:331.20-10.fc20 
will be erased
---> Package kmod-nvidia-3.12.5-302.fc20.x86_64.x86_64 
1:331.20-10.fc20.1 will be erased
...

 > Is that really a good thing?
IMO, this behavior is inacceptable and disqualifies dnf from being made 
distribution-wide default.

> Should we not spare the
> running kernel?

If you look closer, yum doesn't only spare the running kernel, but 
allows a configurable number of multiple versions of some packages 
(notably: kernels and kernel-modules).

The rationale for this is keeping fallback-kernels on the system in case 
a kernel update does not boot or is mal-functioning.

> Or is there some rationale behind this that I am missing?
I think the dnf developers' are missing an important piece of yum history.

Ralf



More information about the users mailing list