[Testing] Final review for Power Management Requirements, Trial-2

[Jim Gettys]

On Mon, 2007-07-16 at 10:33 -0400, Kim Quirk wrote:
> Everyone in the TO: list,
> Can you please review these requirements and send a reply to this
> email that you agree or you want modifications. This is the only
> 'formal' process we have for getting the requirements written down and
> agreed on. 
> 
> It would be great to get this reviewed today -- since there are a
> number of people who depend on understanding how this is supposed to
> work.
> 
> Thanks!!
> Kim
> ------
> 
> Power Management Requirements
> Most of the power management is behind the scene and should be
> invisible to the student. 
> 
> DCON mode

We need different nomenclature.  What I've been calling "dcon mode" is
the system display under control of the dcon, but the CPU powered up.
It is end-user essentially invisible, as the X server and driver
conspire to do this behind the scenes.

The point of this state is to allow us to power down the video unit in
the Geode, which is busily reading memory to drive the display.

I'm happy to call it something else, but we all need to agree.

I've been using the term "ebook mode" for aggressive power management
(whether the machine is in physical ebook mode or not), where the system
suspends whenever it is idle, leaving the screen on (using the dcon).
Ebook mode does not require the hardware to be physically in ebook
configuration.

We don't yet have code to do "ebook mode" automatically; the tests
Walter and I have done to date were done by xbook explicitly suspending
the machine whenever it goes idle.

> 
>       * PW1 Whenever possible the laptop should go into DCON mode,
>         which is the lowest possible power usage to conserve battery
>         life. If the laptop is powered up and left running without
>         intervention with default 'suspend' power settings, the laptop
>         battery life should exceed 24 hours. v

Please keep lifetime numbers out of this document.

The power consumption will be the power consumption, and we are way too
late to change much in the design.  Numbers get written in stone and
come back to haunt you.

>               * PW1.1 While in DCON mode or when coming in and out of
>                 DCON mode, the student will not notice anything as the
>                 screen will remain on. (See Screen off below, which
>                 WILL be noticed by the student).

OHM will dim the backlight automatically.  How quickly it should do so
is TBD.
> 
> Screen off
> 
>       * PW2 After 10 minutes of inactivity through keyboard, mouse,
>         directional, or gamepad keys, the screen will turn completely
>         off (black). 

This "ten minute" number should be a TBD.  We don't know the right set
point yet.  This is a UI choice: someday there might want to be a well
hidden UI for it, but we'll need to select a good default value.

>       * PW3 The screen will turn back on at the press of any key on
>         the keyboard, mouse, directional or gamepad keys. The back
>         light will come back on to the level it was before the screen
>         saver turned on.

Yes.

> 
> Suspend/Resume Button
> 
>       * PW4 Pushing the 'Suspend/Resume' button (used to be the power
>         button), will put the laptop into suspend mode, which includes
>         Screen off. 
>               * PW4.1 In suspend mode, the laptop will continue to
>                 forward wireless packets if it is in the path between
>                 another laptop and an internet connection (school
>                 server, XO as MPP).
>               * PW4.2 Pushing the 'Suspend/Resume' button when the
>                 laptop is suspended will return it to powered on state
>                 with the same activities running as before it
>                 suspended.

Any key, button or keypad touch will wake up the system.

The rational needs to be stated: the point of this is so that if you
want to save as much power as you can while being a "good mesh citizen",
we need a way to suspend the machine and leave wireless on: the child
may have adjusted their antennae, and we want them to be good citizens.
Closing the lid sucks: it unadjusts the antennae, and, if the laptop is
on the ground, lowers them to the ground where they are least effective.

Making it a button the screen would be possible, but is problematic, and
cumbersome.  The screen button might not be visible if it were on the
frame, and the UI interaction is involved enough that the reaction of
the child might be to shut the machine down; thereby weakening the mesh.

You could say that an idle machine should do this (and it will as soon
as we have an idle governor!): but the issue is you can't know for sure
if something is running in the background that the child is really
"done" with the machine and it should save power.  The button will
actually force a suspend (leaving wireless on).


>       * PW5 When the battery gets to its defined shut down level,
>         whether in suspend mode or full operational mode, it will shut
>         down gracefully, without data destruction.
>       * PW6 The student must be able to execute a full power down of
>         the laptop from the Sugar home view.
> 
> From: http://wiki.laptop.org/go/Requirements
> 
> Kim
-- 
Jim Gettys
One Laptop Per Child