Re: HP9000/425e support at Open Source unConference 2014 Kagawa

On 22 April 2014 13:13, Izumi Tsutsui <[email protected]> wrote:
>
> There was also a booth of the "Opencocon" project at the conference.
> The Opencocon is a Linux based distribution optimized for Thin clients:
>  http://opencocon.org/about-cocon
>  https://twitter.com/shimadah/status/439237546912935936/photo/1
>  (the picture is a booth of the opencocon at OSC 2014 Tokyo)
>
> The target machines of the opencocon are Windows 9x generation x86 PCs.
> For example, it can show the Win 8.1 panel on the Toshiba Libretto 50:
>  https://twitter.com/shimadah/status/453539099417133056/photo/1
> The Libretto has Pentium 75MHz CPU, 32MB RAM, and 24bpp graphics.
>
> On the other hand, most m68k machines are >5 year older than
> such Win 9x generation machines.  Most m68k machines have
> ~25MHz CPU, 8-64MB RAM, and only 1, 4 or 8bpp graphics
> (except some 060 monsters).
>
> On such machines even starting only Xserver is heavy enough.
> (It could cause thrashing to move mouse cursor on 12MB X68030)
> Furthermore, there are few X11 applications that supports less
> than 8 bpp graphics (Modern GTK2 apps have some trouble even on 8bpp).
>
> That's the reason why mlterm-fb with sixel graphics is useful
> for demonstration of my m68k machines on these events.

I wonder how feasible it would be to have rdesktop or a vnc client
write directly to a wsfb() type framebuffer, or even via SDL to the
same. Might make an '040 just about usable for a remote display.
Certainly if I was able to use a 4MB 15Mhz 68020 based sun/50 as an X
terminal we must be able to manage *something* :)

>> (When demoing X on a sparc 1 I always found xkobo a nice game to show,
>> I wonder if an '030 or '040 would be fast enough...)
>
> For 1 bpp or 4 bpp machines, pkgsrc/games/xmris and pkgsrc/x11/xearth
> are also good to show. Xeyes is also fun for children.

Don't forget an xneko plus a couple "oneko -toro" processes at
different speeds :)

> If you can prepare any presentation slides, it's also fun to show
> "a machine playing presentation itself":
>  http://movapic.com/pic/201104161105244da8f96468241
>  http://www.ceres.dti.ne.jp/tsutsui/osc2010kobe/img00.html
> This is "making of SMC_TT Ethernet" presentation on TT030 itself
> at OSC Kansai Kobe.
>
> However, Twitter client is still a good demonstration because
> it's easy to prepare (no contents are necessary if the machine can
> run a client application), all visitors can see what it's running,
> and they can also see their own tweets on the demonstrated machine.
>  http://www.youtube.com/watch?v=NytIaugKA8M

Random thought - how much extra load would a twitter bot take? Each
machine could have its own twitter account and respond to certain
commands - uptime, df, uname, first part of dmesg. For bonus (and very
very very slow) option could even tweet a bitmap for dmesg :-p

> Anyway, demonstration at these events is a good motivation
> to build and test various applications on these vintage machines.
>
> Many NetBSD kernel geeks often satisfy if their kernel reaches
> upto multiuser prompt (or even just it prints kernel copyright message),
> but on the events more visible and amazing applications are required
> because the operating system is an implicit software for ordinary users.
>
> Actually PRs pkg/47941, pkg/pkg/48608, and pkg/48672 etc. were
> found on preparing binaries for the demonstrations.
> We are calling it "the event deadline driven development." ;-)

Deadlines are quite amazing productivity devices :)