Mapping form factor

[David Zeuthen]

On Thu, 2009-07-02 at 16:43 +0100, Richard Hughes wrote:
> So, would a "is-laptop" property make more sense? 

No, no, no. You are conflating policy with the form-factor here. You are
trying to be "helpful" but not realizing it's having the opposite
effect. That's what I've been trying to say all along.

To repeat myself: it is _fundamentally_ wrong to expose such properties
because it leads to bad software. Software that tries to be "clever"
about what behavior should be the default by making decisions for the
poor user. Software that acts and feels unpredictable.

The way you need to design your software is

 a) use standard configuration systems like e.g. GConf

 b) document the configuration options so OS installers and OS
    vendors can tweak it the _same_ way they customize other aspects
    of the OS (e.g. spatial- vs browser-based file manager)

Anyway, my view is this: by exposing such properties you are basically
making it very hard to figure out what's going on because people end up
writing "clever" software like that.

> In this way we can
> stop talking about "Computer is low" and start talking about "Laptop
> is low" in the UIs.

You are _guaranteed_ to get things wrong here if you do this - for
example how do you expect to deal with virt where the hypervisor might
give you DMI data that is decidedly wrong (it does this to avoid
breaking other software)?

Just use the word "System" (since even "Computer" is wrong in the virt
case).

Richard, I know you are just trying to push the envelope here and make
things work out of the box - however, your assumption that a "form
factor" value is available and trustworthy is just fundamentally flawed.

    David