* Axel Beckert [Thu Dec 13, 2012 at 03:49:55PM +0100]:
On Thu, Dec 13, 2012 at 02:29:43PM +0100, Markus bionix Ulrich wrote:
we have the bts issue 1220 [1] and mika said to the topic, that maybe it's better to add the package ncurses-term to GRMLBASE instead to GRML_FULL.
[...]
What is your opinion to the topic?
My first thought was "Oh, yes, sure! I explicitly install that on all my boxes, too." But when I thought about why I install it on all my boxes, I noticed that this may be suboptimal for a live CD:
screen has some interesting behaviour depending on where ncurses-terms is installed or not. See http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=694178#10 for the details.
Summary: If screen is started on the e.g. console (TERM=linux) of a box with ncurses-terms installed (contains definitions for screen.linux), it will notice the availability of screen.$TERM definitions and use them. It just uses "screen" otherwise.
If you then SSH from inside the screen to another box where ncurses-term isn't installed, the remote box won't know about the used $TERM (screen.linux) and will argue about it.
So you can't run into that issue at all if ncurses-terms is not installed on the machine where you start SSH inside a screen session.
In that light, I'm not sure if that would add more value than it would cause issues.
Thanks for providing this useful information, Axel! So can we conclude that we don't want to add ncurses-terms to any GRML* software package fail (AKA "don't ship it therefore" :))?
regards, -mika-