please remove update-motd from default install

[Tormod Volden]

Binary package hint: update-motd

update-motd is currently pulled in by update-notifier-common. However, update-motd is totally useless for normal desktop users. Most users do not log in to their computer remotely or on a text console, so any information in /etc/motd is wasted. The current implementations of update-motd (cron jobs running every 15 minutes or a python daemon) are questionable to say the least and tells me the idea of update-motd is not very mature and should not be unleashed on Ubuntu user installations.

[Tormod Volden (tormodvolden)]

Sorry for being so harsh, but I wonder how this ended up in main anyway, and if quality requirements are relaxed when things come from Canonical's engineers or suit their commercial offerings?

[Dustin Kirkland  (kirkland)]

I think you are being rather harsh.

update-motd solved an interesting problem... Actually making the MOTD
dynamic enough to really become a "message of the day". Or "message
of the hour". Or "message of the last 10 minutes". /etc/motd in UNIX
has been around for almost 40 years, and no OS that I know of has
instituted a framework at the distribution level for allowing an
administrator to dynamically build the MOTD.

The original implementation was hacked out just in time for Feature
Freeze for Intrepid. There were no corners cut. It went through the
normal process for inclusion. The development runway was short, and
the 10-minute cronjob was the easiest to implement in the given
timeframe.

Trust me, I was *never* happy with the cronjob design, and have been
looking for a better solution, albeit in my spare time, as I'm not
tasked to work on update-motd.

Over the weekend, I rewrote update-motd from scratch, replacing the
cronjobs with a single daemon. In my testing, the daemon was
extremely fast, and efficient, and had nearly zero impact to a running
system. That said, some people did raise the issue (in a far more
constructive manner than yourself) that they didn't see the point of
the daemon. That this data should be collected at login.

I spent another day re-writing it yet again to use /etc/profile.d.
Which, again, was not acceptable to another party.

And so I just completed a fourth rewrite, integrating it into
pam_motd. This is the first implementation that I've actually been
happy with. And it took some time, and a number of constructive
conversations to get there.

Soon enough, you'll get your wish, and the update-motd package will be
eliminated as its functionality is subsumed directly into PAM.

Next time, though, you should try a less harsh approach. Come with a
proposed solution, or code in hand. Your request will be met with
much more consideration.

Dustin

[to be removed (liw)]

Random historical note: Back when I was young...er, motd was a file that root was meant to edit by hand. It was for messages from the sysadmins to the users, not from random system services to users that happen to log in via a terminal. There were some other mechanisms for that kind of thing, such as wall, e-mail, and custom site-specific things. Those other mechanisms have since broken down, pretty much, so perhaps an automatically updated motd is a necessary half-subsitute.

[Tormod Volden (tormodvolden)]

OK, I was too harsh and exaggerated a bit. The rant was building up from a feeling of seeing similar things going on before and it was wrong to let it all out here. And I am not sour at you for developing this, but at those who let this into Intrepid, so please do not take this personally. Other than I would like to encourage thinking over what Ubuntu needs and not, and a server guy will have different perspectives than the average Ubuntu user. But I am not telling you what you should work on. And I understand that Canonical needs to make money, be it by promoting landscape solutions and what not.

This does not change my opinion that update-motd is really a bad idea.

> update-motd solved an interesting problem

No, this is not interesting for normal users. Even for multi-users systems, I am sure people don't read the motd even if they would happen to log in on a console. An administrator will have his own status scripts and routines to pick up this information.

> and had nearly zero impact to a running system.

This is a general problem which added to the fire: Developers are testing on their great computers (or Dell 10v) and we end up with things like for instance apt-xapian-update or scrollkeeper-update which totally destroy the user experience for those with more moderate computers. Or another daemon here and python applet there, and that's why an empty desktop needs 1GB of RAM. All these nearly-zero-impact things add up (or multiply even).

> Come with a proposed solution, or code in hand.

That is what I did in the bug report: Remove the Recommends from update-notifier common to make sure it is not included in a default install. If some server guy wants to install it (from universe), fine.

I am glad to see this is developing in a more sane direction with pam-motd, but again I don't see why this is rushed into main or default installs before maturing a bit.