future of smart

[Jeff Johnson]

On Jun 11, 2008, at 6:29 PM, Grant McWilliams wrote:

> I found the sheer volume of responses to my post to be pretty
> encouraging.
>
> There's a lot of us using it that are too busy to request or work  
> on features. I have it on 100 machines around the world and was  
> going to put some time in to help reorganize the docs so they made  
> more sense and were easier to find and read but time is something  
> that never seems to materialize. :-(
>

Hint: developers have a busy "real world" life too.

> I do think some of the features on my consolidated list need to be  
> added though in order for SmartPM to remain viable. The mirrors  
> issue and GPG key checking are both "musts". And then I still have  
> my wishes... :-)
>
> I'd like to see Smart become THE backend for Packagekit. Currently  
> I have some reservations though. More than one time I've gotten  
> SmartPM into an upgrade/downgrade loop that would never stop. I  
> think every time it was a packagers problem because it went away a  
> couple days later but explain that to the users that don't  
> understand why they can update their machines. Solution? I don't know.

I'm not sure that being THE back end for PackageKit is really useful.

PackageKit is just some plumbing to split GUI <-> engine, and the
code is changing too rapidly to seriously assess usefulness yet imho.

FYI  depsolver performance & behavior were fairly carefully analyzed  
by the
EDOS project in 2005:

      http://www.mancoosi.org/edos/manager.html

Even in 2005 smart was arguably "best-of-breed", and far superior
to apt-rpm, Gustavo's former child. ;-)

But 3 years later, there are new candidates around, including
Conary and OpenSolaris IPS and SuSE's spiffy new zypp tool.

Hard to tell whether package management even matters any more.

However, my bet is still on smart. Stabilizing a package manager in the
wild is lots lots harder than it looks.

73 de Jeff