<?xml version="1.0" encoding="UTF-8"?>

<rdf:RDF 
  xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
  xmlns="http://purl.org/rss/1.0/"
  xmlns:dc="http://purl.org/dc/elements/1.1/"
  xmlns:syn="http://purl.org/rss/1.0/modules/syndication/"
>

  <channel rdf:about="http://lwn.net/headlines/240399/">
    <title>LWN: Comments on "Package management in Gentoo Linux"</title>
    <link>http://lwn.net/Articles/240399/</link>
    <description>
This is a special feed containing comments posted
to the individual LWN article titled &quot;Package management in Gentoo Linux&quot;.

    </description>

    <syn:updatePeriod>hourly</syn:updatePeriod>
    <syn:updateFrequency>2</syn:updateFrequency>
    <items>
      <rdf:Seq>
	<rdf:li resource="http://lwn.net/Articles/243360/rss" />
	<rdf:li resource="http://lwn.net/Articles/242270/rss" />
	<rdf:li resource="http://lwn.net/Articles/241730/rss" />
	<rdf:li resource="http://lwn.net/Articles/241647/rss" />
	<rdf:li resource="http://lwn.net/Articles/241528/rss" />
	<rdf:li resource="http://lwn.net/Articles/241190/rss" />
	<rdf:li resource="http://lwn.net/Articles/240953/rss" />
	<rdf:li resource="http://lwn.net/Articles/240881/rss" />
	<rdf:li resource="http://lwn.net/Articles/240866/rss" />
	<rdf:li resource="http://lwn.net/Articles/240783/rss" />
	<rdf:li resource="http://lwn.net/Articles/240757/rss" />
	<rdf:li resource="http://lwn.net/Articles/240733/rss" />
	<rdf:li resource="http://lwn.net/Articles/240686/rss" />
      
      </rdf:Seq>
    </items>

  </channel>
    <item rdf:about="http://lwn.net/Articles/243360/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/243360/rss</link>
      <dc:date>2007-07-27T11:45:01+00:00</dc:date>
      <dc:creator>fintan</dc:creator>
      <description>
      I think conary is more along the right approach. &lt;a rel=&quot;nofollow&quot; href=&quot;http://wiki.rpath.com/wiki/Conary&quot;&gt;http://wiki.rpath.com/wiki/Conary&lt;/a&gt;&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/242270/rss">
      <title>AUFS for Transactional Upgrades</title>
      <link>http://lwn.net/Articles/242270/rss</link>
      <dc:date>2007-07-19T15:48:44+00:00</dc:date>
      <dc:creator>ferringb</dc:creator>
      <description>
      Actually I tried something similar a while back; unionfs sandboxing of the phases to try and get the ability to truly track/reverse what pre_inst/pre_rm were upto, and track ebuilds builds where userpriv restrictions were in effect; problem I had was that it always wound up making gcc spew in a non-obvious way for compilation.&lt;br&gt;
&lt;p&gt;
Either way, interesting to see someone playing with it still (nature of some of the phases, it's kind of required long term imo).&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/241730/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/241730/rss</link>
      <dc:date>2007-07-14T09:10:08+00:00</dc:date>
      <dc:creator>tres</dc:creator>
      <description>
      Not as mush as Paludis' main developer, Ciaranm did.  That guy has somewhat of an abrasive personality (to say the least).&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/241647/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/241647/rss</link>
      <dc:date>2007-07-13T14:55:07+00:00</dc:date>
      <dc:creator>muwlgr</dc:creator>
      <description>
      That very Paludis that pushed DRobbins back out of Gentoo :&lt;br&gt;
&lt;p&gt;
&lt;a rel=&quot;nofollow&quot; href=&quot;http://www.google.com/search?q=robbins+gentoo+paludis&quot;&gt;http://www.google.com/search?q=robbins+gentoo+paludis&lt;/a&gt;&lt;br&gt;
&lt;a rel=&quot;nofollow&quot; href=&quot;http://lwn.net/Articles/224082/&quot;&gt;http://lwn.net/Articles/224082/&lt;/a&gt;&lt;br&gt;
&lt;a rel=&quot;nofollow&quot; href=&quot;http://lwn.net/Articles/224615/&quot;&gt;http://lwn.net/Articles/224615/&lt;/a&gt;&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/241528/rss">
      <title>AUFS for Transactional Upgrades</title>
      <link>http://lwn.net/Articles/241528/rss</link>
      <dc:date>2007-07-12T20:19:08+00:00</dc:date>
      <dc:creator>hathawsh</dc:creator>
      <description>
      For near-transactional upgrades, consider an aufs-based chroot.  (Note that this idea applies equally well to any package manager.)  Here is someone who tried it:&lt;br&gt;
&lt;p&gt;
&lt;a rel=&quot;nofollow&quot; href=&quot;http://blog.vrplumber.com/1889&quot;&gt;http://blog.vrplumber.com/1889&lt;/a&gt;&lt;br&gt;
&lt;p&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/241190/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/241190/rss</link>
      <dc:date>2007-07-10T16:00:20+00:00</dc:date>
      <dc:creator>rise</dc:creator>
      <description>
      If you're not using static binaries anywhere in the process (a big if)
a &lt;a
href=&quot;http://asic-linux.com.mx/~izto/checkinstall/&quot;&gt;CheckInstall&lt;/a&gt;/installwatch-style
solution might work.  Basically it uses a library preload to catch all
file accesses and redirect them to a temporary area while overlaying
the results over the actual filesystem.  Then it bundles up all the
changes it saw into a package.  This is a nice but sub-optimal
solution for software lacking true packages, though I use it heavily
to make random source-compiled software trackable and uninstallable.
However it should work nicely for transactions - just delay committing
the overlay until the process completes properly.

      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/240953/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/240953/rss</link>
      <dc:date>2007-07-08T20:29:21+00:00</dc:date>
      <dc:creator>dirtyepic</dc:creator>
      <description>
      &lt;i&gt;3. Transactional upgrades. If you want to upgrade a slew of software,
merge all the files into a temporary holding directory and wait until all
packages and their dependencies have successfully compiled before
updating the live system. Having to chase down build failures in the
middle of an &quot;emerge&quot;, when your system is currently in a broken state,
is irritating.&lt;/i&gt;&lt;br&gt;
&lt;br&gt;
Interesting idea, but I'm not sure how that would work or why introducing massive changes to the system all at once rather than incrementally would help anything.  Would you link to the system libraries or to the ones you've just built in the holding area?  What happens when those libraries are suddenly relocated or overwritten?&lt;br&gt;
&lt;br&gt;
The best way to handle updating is one package at a time.  If something breaks, then you only have to deal with that package.  Blindly running emerge world is usually what gets people into trouble in the first place.&lt;br&gt;
&lt;br&gt;
&lt;i&gt;4. A better etc-update. The one that is included should be taken out back and shot :P&lt;/i&gt;&lt;br&gt;
&lt;br&gt;
I don't honestly know why it's still around and the default.  dispatch-conf forever.
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/240881/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/240881/rss</link>
      <dc:date>2007-07-06T19:56:31+00:00</dc:date>
      <dc:creator>dberkholz</dc:creator>
      <description>
      Great suggestions!&lt;br&gt;
&lt;p&gt;
- It needs to do more than just ldd, so it can handle all types of languages. For example, Perl or Python modules need to get handled a bit smarter. Various people have worked a little on this problem, but nobody in Gentoo has done a good job of finishing it.&lt;br&gt;
&lt;p&gt;
- The ebuild cache as it exists now is a little subpar. You've got the current ebuild in /var/db/pkg/, or you can look in the CVS Attic via your anoncvs checkout or &lt;a href=&quot;http://sources.gentoo.org/&quot;&gt;http://sources.gentoo.org/&lt;/a&gt;.&lt;br&gt;
&lt;p&gt;
- I really like the transactional idea.&lt;br&gt;
&lt;p&gt;
- Some other possibilities do exist for updating your config files such as dispatch-conf, conf-update, cfg-update (all of which are part of Portage itself or in the main Portage tree) or the new etc-proposals (sunrise overlay). Try 'em out.&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/240866/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/240866/rss</link>
      <dc:date>2007-07-06T16:39:13+00:00</dc:date>
      <dc:creator>cventers</dc:creator>
      <description>
      What I think would be really neat is if these package managers had the &lt;br&gt;
following:&lt;br&gt;
&lt;p&gt;
1. Dynamic library tracking, because revdep-rebuild SUCKS! The package &lt;br&gt;
manager knows what binaries it is installing into the live system, it &lt;br&gt;
should be able to 'ldd' and remember from then on. The hope, of course, &lt;br&gt;
is that the package manager would be smart enough to recompile any &lt;br&gt;
packages dependent on the library.&lt;br&gt;
&lt;p&gt;
2. Ebuild cache... periodically, Gentoo deletes ebuilds from the portage &lt;br&gt;
tree. The problem is that you may have an old version of some package &lt;br&gt;
that Gentoo no longer supports which depends on a library you want to &lt;br&gt;
upgrade. If you still have the source, Gentoo should happily rebuild the &lt;br&gt;
old (no longer supported) version of the software against the new &lt;br&gt;
library.&lt;br&gt;
&lt;p&gt;
The lack of this ability leads to occasional frustration when you have to &lt;br&gt;
upgrade a library due to a security vulnerability, only to discover that &lt;br&gt;
you now have to upgrade other packages just because Gentoo deleted the &lt;br&gt;
ebuild for the version you were using.&lt;br&gt;
&lt;p&gt;
3. Transactional upgrades. If you want to upgrade a slew of software, &lt;br&gt;
merge all the files into a temporary holding directory and wait until all &lt;br&gt;
packages and their dependencies have successfully compiled before &lt;br&gt;
updating the live system. Having to chase down build failures in the &lt;br&gt;
middle of an &quot;emerge&quot;, when your system is currently in a broken state, &lt;br&gt;
is irritating.&lt;br&gt;
&lt;p&gt;
4. A better etc-update. The one that is included should be taken out back &lt;br&gt;
and shot :P&lt;br&gt;
&lt;p&gt;
Gentoo is great, but in some ways I feel like it is just the tallest &lt;br&gt;
midget. I really wish I had the time to help on the code, because I feel &lt;br&gt;
that these features would greatly enhance the OS. A guy can dream, can't &lt;br&gt;
he?&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/240783/rss">
      <title>How painful is the migration?</title>
      <link>http://lwn.net/Articles/240783/rss</link>
      <dc:date>2007-07-06T00:43:33+00:00</dc:date>
      <dc:creator>dberkholz</dc:creator>
      <description>
      I mentioned that the latest paludis releases have compatibility with Portage config files. Try USE=portage emerge paludis, and see how things go. It should pretty much work.&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/240757/rss">
      <title>How painful is the migration?</title>
      <link>http://lwn.net/Articles/240757/rss</link>
      <dc:date>2007-07-05T20:01:47+00:00</dc:date>
      <dc:creator>smitty_one_each</dc:creator>
      <description>
      There is a shell script that worked great for me going portage-&amp;gt;paludis on the paludis.pioto.org site.&lt;br&gt;
I have no experience running the two in parallel, but I'll venture that you'll move away from emerge rather speedily. ;)&lt;br&gt;
After switching to paludis, I installed layman, and used it to kick-start some of the overlays.gentoo.org stuff.  Paludis expects more discipline, apparently, and there was some manual effort to configure, say, the emacs overlay.  Had to create several files and directories to silence the error traces.&lt;br&gt;
I think there is some provision for using the old-style emerge configuration files, but paludis again does a better job of herding the cats into a few coherent files.&lt;br&gt;
Hats off to emerge.  It had a great run.&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/240733/rss">
      <title>How painful is the migration?</title>
      <link>http://lwn.net/Articles/240733/rss</link>
      <dc:date>2007-07-05T18:33:03+00:00</dc:date>
      <dc:creator>felixfix</dc:creator>
      <description>
      I have just started an emerge of paludis and will read up on it, but maybe you can  answer two quickie questions?&lt;br&gt;
&lt;p&gt;
How painful is switching to Paludis?&lt;br&gt;
&lt;p&gt;
Is it possible to run Paludis in parallel with portage, and what is the pain level?&lt;br&gt;
      
      </description>
    </item>
    <item rdf:about="http://lwn.net/Articles/240686/rss">
      <title>Package management in Gentoo Linux</title>
      <link>http://lwn.net/Articles/240686/rss</link>
      <dc:date>2007-07-05T12:27:27+00:00</dc:date>
      <dc:creator>smitty_one_each</dc:creator>
      <description>
      Great article.&lt;br&gt;
Venerable though portage may be, the nail in its coffin is going to be benchmarking.  As a paludis user, I can tell you that dependency resolution is _a lot_ faster with McCreesh's solution than with portage, and I won't hazard a guess about accuracy.&lt;br&gt;
      
      </description>
    </item>
</rdf:RDF>

