DNF Any testing time-frame?

Ales Kozumplik akozumpl at redhat.com
Thu Jun 21 11:29:59 UTC 2012

On 06/20/2012 01:04 PM, tim.lauridsen at gmail.com wrote:
> Would it not be better to work with yum upstream to make the current yum
> depsolver more modular so you could plugin another libsolv based
> depsolver, instead of making a fork of yum and starts trashing the
> current API. There is a lot more to yum, that just solving dependencies.
> And making a fork there is not fully compatible will put a lot of work
> on your shoulders :) without the benefit on the work done by yum
> upstream :) like parallel download etc.
> Tim

Hello Tim,

Be assured I am in contact with Yum developers and the new features 
happening for F18 there are planned to be integrated to DNF.

The decision to fork yum into using libsolv instead of trying to evolve 
it slowly was a difficult one yet it was the right one. Unified 
depsolving is only a part of the project, the other primary goal is 
arriving at concrete, cleaned up API for external applications and 
plugins. This is very hard to get done without having free hands to 
refactor, remove, cleanup and change for better testability because in 
Yum one always has to look behind his back for tricky backward 
compatibility issues.

I think the best way to describe the "DNF" project really is as "the 
next-gen yum". Departing from the old APIs is a part of it. There are 
little alternatives also: the gradual deprecation of some of the Yum's 
legacy interfaces hasn't been very successful in the past. Or I could 
keep DNF under the lid for another one or two Fedora releases, but I 
find it a better alternative to package this early version: both for 
those Fedora users who will bravely try it (and hopefully report 
feedback) and the community members who might join the effort.


More information about the devel mailing list