Re: [tlug] Running without Gnome/KDE/xfce/whatever. (was: Ubuntu 16.04-LTS Japanese Text Input)

["Stephen J. Turnbull" <turnbull@example.com>]

Attila Kinali writes:

 > Well.. my setup is from a time, where Xfree86 3.3.6 was HOT!

Good for you!

 > (I wonder why i can even remember the version?) and these
 > nice wrappers didn't exist yet.

Actually, the Debian wrappers go back to at least that time, but they
were very simple then (and they may actually have been part of the
XFree86 distribution rather than Debian-contributed).  The most
important thing they did was handle xauth stuff.  I know that they did
that much, because I never did learn to do that by hand. :-)

 > Being self-contained or at least having a low number of dependencies
 > is not on-vogue anymore. Instead authors make many assumptions about
 > how your system looks like and lable it as user error when it those
 > assumptions do not hold. :-(

Yup.  It makes systems more complex and fragile.  I thought we'd
learned something from the early stages of GNU/Linux/FLOSS[1]
development, but apparently not.

 > I cannot find that part in the ICCCM (definite lack of chocolate)
 > but i was sure it was a clear MUST NOT...

Even if so, you need to be careful about what these requirements mean
and where they apply.  The ICCCM is specifically about *interclient*
communication, so doesn't apply to communication with non-X apps, and
it also doesn't apply to X clients that are communicating about
conditions that are necessarily local to a given host.  For example,
you could imagine a situation where a file manager provides both
change notification services via dbus and a GUI interface for the
user.  Then a program like Xpdf could be a client of the dbus
notification service but it's not really communication between X
clients.  (I wouldn't think much of the architecture of that file
manager, but that's another question.)

Also, I seem to recall a footnote in the ICCCM that says that ICCCM
transfers can choose to use efficient transports (similar to the way a
Unix socket rather than a TCP/IP stream may be used for the X
connection itself), to avoid forcing the X server to receive, store,
and retransmit large amounts of data.  dbus would be counted as an
"efficient transport" in many contexts.


Footnotes: 
[1]  I've been reading Moody's *Rebel Code*, and it's interesting to
note the evolution of these ideas on simplicity, modularity, and
minimal assumptions in the early FLOSS movement.  More about that in
another post.