On Debian discussions

In my article „Last time I’ve used network-manager“ I made a claim for which I’ve been criticized by some people, including Stefano, our current (and just re-elected) DPL.  I said that a certain pattern, which showed up in a certain thread, were a prototype for discussions in the Debian surroundings.

Actually I have to commit, that this was a very generalizing statement, making my own point against the discussion point back directly at myself.
Because as Stefano said correctly there has been some progress in the Debian discussion cult.
Indeed, there are examples of threads, were discussions followed another scheme.
But to my own advocacy I have to say that such changes are like little plants (in the botanical sense). They take their time to grow and as long as they are so very new, they are very vulnerable to all small interruptions. Regardless of how tiny those interruptions may seem.

I’ve been following Debian discussions for 6 or 7 years. That scheme I was describing was that which had the most visibility of all Debian discussions. Almost every discussion which were important for a broader audience followed that scheme. It has a reason that Debian is famous for flamewars.
In a way its quiet similar to the network-manager perception, some people have. Negative impressions manifest themselves. Especially if they have years of time.
Positive impressions does not have a chance to manifest themselves as long as the progress is not visible enough to survive small interruptions.

I hope that I didn’t cause to much damage with my comment, which got cited (context-less) on other sites. Hopefully the Debian discussion cult will improve further to a point where there is no difference between the examples of very good, constructive discussions we already have in some parts of the project and the project-wide decision-making-discussions which affect a broad audience and often led to flamewars.

Directory-dependent shell configuration with zsh (Update)

For a while I’ve been struggling with a little itch. I’m using my company notebook for company work and for Debian related stuff. Now, whenever I switch between those two contexts, I had to manually fix the environment configuration. This is mostly related to environment variables, because tools like dch et cetera rely on some, which need to be different for the different contexts, like DEBEMAIL.
A while ago I had the idea to use directory dependent configuration for that purpose, but I never found the time and mood to actually fix my itch.
Somewhere in the meanwhile I applied a quick hack („case $PWD in …; do export…; esac“) to my zsh configuration to ease the pain, but it still did not feel right.


For the impatient: Below you find a way to just use whats described here. The rest of the article just contains detailed information on how to implement something like this.


The other day I were cleaning up and extending my zsh configuration and it came to my mind again. I then thought about what my requirements are and how I could solve it. First I thought about using a ready solution, like the one in the Grml Zsh configuration, but at that point I did not remember it (it needed a hint by a co-worker *after* I finished the first version of my solution). Then I came up with my requirements:

  • Separate profile changing logic from configuration (as far as possible):I don’t want to re-dive into a script logic every time I decide to change something, like adding a variable or changing it. Generally I find a declarative approach much cleaner.
  • Avoid repeating myself
    Basically all I do when switching profiles is to change environment variables. Usually I don’t want my shell to do extraordinary things, like brewing coffee when I switch the context, so I’d like to avoid typing an „export foobar..“ for every single environment variable and every single profile.
It lead to a configuration approach as a first start. When thinking about how to represent the configuration I looked into the supported data types in zsh. zsh supports arrays, which is perfect for my need. I came up with something like that:

typeset -A EMAILS ENV_DEBIAN ENV_COMPANY
EMAILS=(
  „private“     „foo@bar.org“
  „company“     „baz@foo.org“
  „debian“      „schoenfeld@debian.org“
)
ENV_DEBIAN=(
  „DEBEMAIL“  „$EMAILS[debian]“
  „EMAIL“     „$EMAILS[debian]“
 )
ENV_COMPANY=(
  „DEBEMAIL“  „$EMAILS[company]“
)

The next part was selecting the right profile. In the first version I used the old case logic, but it was breaking my separate logic and configuration paradigm. Approximate at this point the co-worker lead me to the grml approach, which I borrowed an idea from:


# Configure profile mappings
zstyle ‚:chpwd:profiles:*company*‘ profile company
zstyle ‚:chpwd:profiles:*debian*‘ profile debian

and the following code to lookup profile based on $PWD:


1 function detect_env_profile {
2   local profile
3   zstyle -s „:chpwd:profiles:${PWD}“ profile profile || profile=’default‘
4   profile=${(U)profile}
5   if [ „$profile“ != „$ENV_PROFILE“ ]; then
6   print „Switching to profile: $profile“
7   fi
8   ENV_PROFILE=“$profile“
9 }

For an explanation: zstyle is a zsh-builtin which is used to „define and lookup styles“, as the manpage says, or put different: Another way to store and lookup configuration values.
Its nice for my purpose, because it allows storing patterns instead of plain configuration values which can be compared against $PWD easily with all of the zsh globbing magic. This is basically whats done in line 3. zstyle then sets $profile to the matching zstyle configuration in the :chpwd:profiles: context or to ‚default‘ if no matching zstyle is found.

The (almost) last part is putting it together with code to switch the profile:


1 function switch_environment_profiles {
2   detect_env_profile
3   config_key=“ENV_$ENV_PROFILE“
4   for key value in ${(kvP)config_key}; do
5     export $key=$value
6   done
7}

The only non-obvious part in this are lines 3 and 4. Remember, the profiles were defined as ENV_PROFILE, where PROFILE is the name of the profile. We cannot know that key in advance, therefore we have to construct the right environment variable from the result of detect_env_profile. We do that in line 3 and lookup this environment variable in line 4.
The deciding aspect for that is the P-flag in the parameter expansion. It tells zsh that we do not want the value of $config_key, but instead the value of $WHATEVER_CONFIG_KEY_EXPANDS_TO.
The other flags k and v tell zsh that, from the array, we want both: keys and values. If we’d omitted those flags it would have given us the values only.
We then loop over that to configure the environment. Easy, hu?

We would be finished, if this would do anything. The code above needs to be called. Lucky for us thats pretty easy to achieve, as zsh has a hook for when a directory is changed. Making all this work is simply a matter of adding something like this:


function chpwd() {
  switch_environment_profiles
}

Now, one could say, that the solution in the grml configuration has an advantage. It allows calling arbitrary commands on profile changing, which might be useful to *unset* variables in a given profile or whatever you could think of.
Well, its a matter of three lines to extend the above code for that feature:


# Taken from grml zshrc, allow chpwd_profile_functions()
if (( ${+functions[chpwd_profile_$ENV_PROFILE]} )) ; then
  chpwd_profile_${ENV_PROFILE}
fi

to the end of switch_environment_profiles and now its possible to additionall
y add a function chpwd_profile_PROFILE which is called whenever the profile is changed to that profile.



USAGE: I have put the functions into a file which can be included into your zsh configuration, which can be found on github.
Please see this README and the comments in the file itself on further usage instructions.

password-gorilla 1.5.3.4 ACCEPTED into unstable

The password-gorilla package has lacked some love since a while and at some point in time I orphaned it.
That happened due to the fact, that the upstream author was pretty unresponsive and inactive and my own TCL skills are very limited. As a result password-gorilla package was in a bad state, at least from a user point of view, with several (apparently) random happening error message and alike, stalling feature development etc.

But in the meanwhile there was a promising event arising. A guy, named Zbigniew Diaczyszyn, wrote me a mail that he intended to continue upstream development. Well, meanwhile is kind of an understatement. That first mail already happened in December 2009. And he asked me, if I’d like to continue maintaining password-gorilla in Debian. I agreed to that, but as promising as it sounded to have a new upstream, I was not sure if that would work out. However: My doubt were not justified.

In the time between 2009 and now Zbigniew managed to become the official upstream (with the accreditation of the previous upstream), create a github project for it and make several releases.


I know there are several people out there who tested password-gorilla. I know there were magazine reviews including the old version, which were a bit buggy with recent tcl/tk versions. It made a quiet good multi-platform password manager, with support for very common password file formats, stand in a bad light.
I recommend previous users of password-gorilla to try the new version, which recently has been
uploaded to unstable.

Last time I’ve used network-manager..

Theres an ongoing thread on the Debian mailing lists about making network-manager installed by default on new Debian installations. I won’t say much about the thread. Its just a prototype example for Debian project discussions: Discuss everything to death and if its dead discuss a little more. And – very important – always restate the same arguments as often as you can. Or if its not your own argument you restate, restate the arguments of others. Ending with 100 times stated the same argument. Even if its already disproved.

I don’t have a strong opinion about the topic in itself. However there is something I find kinda funny. A statement brought up by the people who strongly oppose network-manager as a default.
A statetement I’ve heard so often that I can’t count it anymore.

The last time I’ve tried network-manager it sucked.

It often comes in different masquerades, like:

  • network-manager is crap.
  • network-manager is totally unusable
  • network-manager does not even manage to keep the network connection during upgrades
But it basically boils down to that basic essence of the sentence I’ve written above. Sometimes I ask people who express this opinion a simple question:

When did you test network-manager the last time?

The answers are different but again the basic essence of the answers is mostly the same (even if people would never say it that way):

A long time ago. Must have been around Etch.

And guess what: There was a time when I had a similar opinion. Must have been around Etch.
During the life cycle of network-manager between Etch and now a lot has happened. I restarted using network-manager at some point of the Lenny development.
My daily driver for the management of my network connections on my notebook. Yes, together with ifupdown because, yes, network-manager does not support every possible network-setup with all of the special cases possible. But it supports auto-configuring of wired and wireless devices. Connecting to a new encrypted network, either in a WLAN or in a 802.1x LAN, using UMTS devices, using tethering with a smart phone. And everything: on a few mouse-clicks.
Yes, it had some rough edges in that life cycle. Yes, it had that nasty upgrade bug, which was very annoying.
But face it: It developed a lot. Here are some numbers:
Diffstat between the etch version and the lenny version:
 362 files changed, 36589 insertions(+), 36684 deletions(-)
Diffstat between the Lenny version and the current version in sid:
 763 files changed, 112713 insertions(+), 56361 deletions(-)
The upgrade bug has been solved recently. Late. But better late then never.
So what does that mean? It means that, if your last network-manager experience was made with Lenny or even worse around Etch, you should better give it another try, if you are interested in knowing what you talk about. For now it seems that a lot of people do not know. Not even in a distance.