matthew green wrote:
>    David Brownlee wrote:
>    >
>    >         Just tried this out, and it works really well, though I did find a
>    >         couple of small points:
>    >             a) Switching virtual consoles resets the scroll buffer
> 
>    that's not a bug, that's a feature :). There are basically to way
>    to implement console scrolling: using VGA memory to store the buffer
>    (Linux) or allocating KVM memory (FreeBSD). I prefered the second one
>    because I like to scroll back as much as I want (you have no physical
>    limitations). But in order not to wasted to much KVM the scroll
>    buffer is shared among virtual screens. The other way would be to
>    allocate a scroll buffer for each virtual screen but that seem to much
>    a waste. BTW the behaviour is the same as FreeBSD a Linux...
> 
> i don't like this.  i don't care what freebsd or linux do.  this behaviour
> is totally alien to me (a heavy screen user).  at the very least, i'd expect
> the scroll buffer to stay alive across switches, but i'd prefer that each
> VT had it's own scroll buffer, allocated out of KVM (pageable?) as you said.
> 
> .mrg.

The "very least" is too much complex if you think about it (I'd rather not
discuss
about it because that's for sure the worst solution). But I agree with you
that it can be appealing to have a per screen scroll buffer.
I usually set SCROLLBUFFER_DIM to 20,000 (chars) that is a 80x250 virtual
screen:
in the per screen scroll buffer case if you have 4 screens active that would be
approximately 160 KB of KVM (20,000 * 2 (byte per char) * 4). Isn't that too
much?
It would not be a problem anyway to implemet that solution... 
if I could count an you, David and Matthew, for a vote pro code merge I'll do it
right now :)

	Paolo