HAL 0.1 release

[Eric Gillespie]

Joe Shaw <joe at ximian.com> writes:

> With regards to the file format, why did you use tags like:
> 
> <data class="linux">
>    <data class="module">
>       <data class="name">syn53c8xx</data>
>    </data>
> </data>
> 
> over a more structured-feeling format:
> 
> <os type="linux">
>    <module name="syn53c8xx"/>
> </os>

Discover doesn't know about modules.  It doesn't know about
operating systems.  It doesn't know about XFree86.  You can store
any kind of structured data you want in the discover data files.
*That* is why we designed it that way.

> or to borrow a phrase from Havoc, abstraction through
> directories, in which you'd have a linux directory which would
> have a device file with just:

That's exactly how it works.  The data

<data class="linux"
  <data class="module">
    <data class="name">syn53c8xx"></data>
  </data>
</data>

is retrieved with a pathname: "linux/module/name".  The
difference between what you suggest and what we did is that
discover itself doesn't know anything about the data.  It neither
knows nor cares that the first path element is the name of an
operating system.  A module auto-loader (such as the example
script provided in the discover package, the one used by Progeny)
extracts that information for its purposes, an XFree86
configurator (such as the Debconf scripts in Debian's XFree86
packaging) extracts information it understands, and so on.

--  
Eric Gillespie <*> epg at pretzelnet.org