Subject: Re: Patch to add console scrollback support.
To: Lord Isildur <>
From: Greywolf <>
List: tech-kern
Date: 06/01/2001 15:21:37
On Fri, 1 Jun 2001, Lord Isildur wrote:

# On 1 Jun 2001, Johan Danielsson wrote:
# > Then disable it on your 8M machines. It's not gonna waste less memory
# > if it lives in user space.
# theres a fundamental difference between userland and kernel. to shovel
# things into the kernel that really havent a good need (a user not feeling
# like running screen does not in my mind constitute 'need') to be there is
# one of the primary causes of bloat, degradation, and overall reduction in
# quality. any addition or expansion of features in the kernel _should_
# (and _are_) be well-debated before getting in there.

Well, this one's getting debated into the ground.  Again.

But the result seems to be somewhat different:  More people are proponents
of this change actually happening.

You're also missing the points of difference between "screen" and having
the scrollback in the wscons:

* The boot messages from the rc scripts are not logged anywhere, and once
  they scroll off the screen, they're hosed.

* Ditto the kernel boot messages (although they do get shoved into
  /var/run/dmesg.boot (that was a cool idea; thanks to the person who
  did that one!)).

* Not everyone has the luxury of having a second computer by which to
  establish a serial console connection.

* screen does not provide a getty/login, and requires one to log in to
  start it.

* The scrollback buffer code for wscons -- and let me emphasize my mantra
  -- *can* *be* *made* *an* *option*.

Now, let me ask the question again:  Which of these points are people
really having this much difficulty with?   We should have had scrollback
in the wscons driver available for a long time now; the specious arguments
which boil down to "I don't like it" are just insipid and have been done
to death.  If you don't like the memory being wasted, comment it out of
the config.  But this sort of functionality is just too fscking useful
to be left out, and personally, I don't find that much in the coding arena
that I can say this about WRT NetBSD.

[by the way, my comment regarding "we should have had..." above should
 not be taken to mean that the coder(s) ha{s,ve}n't been doing their

# > I'm pretty sick of these types of arguments.
# why? they might sound monotonous after a while, but it is exactly the
# same for someone of the opposite point of view to say 'im pretty sick of
# these continual additions (or proposed additions) of fluff into the
# kernel' and it would be exactly the same assertion.

Fluff.  Weeds.  Both fall into the category of things that are
undesirable.  I know an herbalist who considers roses to be weeds, for
example.  I consider _for my environment_ ipv6 to be fluff -- I comment it
out (my kernel has dropped to half its previous size!) because my
router/firewall, while a nice little box, won't handle IPv6 unless I
tunnel it over ipv4, and it's not worth my while to do that right now.
Ditto CODA, LFS, PORTAL, UMAPFS (though you can probably understand more
why I drop that stuff out of the kernel).

What do you think the reaction of this list would be were I to suggest
that IPv6 takes up unnecessary space in the kernel?  Do you think for a
moment that the suggestion would be met with any form of consideration
beyond "If you don't like it, don't compile it in"?

The same principle applies with wscons+scrollback.

# isildur

NetBSD:  the OS WORM -  Write Once, Run Many.