1 |
On Sunday 17 January 2010 22:14:06 Neil Walker wrote: |
2 |
> Joerg Schilling wrote: |
3 |
> > how do we |
4 |
> > prevent that DeviceKit will become the same desaster as hald? |
5 |
> |
6 |
> The only way to be sure of that is to write your own replacement for HAL. |
7 |
> ;) |
8 |
|
9 |
That might not be a bad idea.... |
10 |
|
11 |
I never agreed with the implementation of hal. An abstract layer sounds good, |
12 |
but why must it abstract ALL hardware? Most software already knows what type |
13 |
of devices it is going to use, so that software should either do it's own |
14 |
abstraction, or a utility library should do it, but be limited to what devices |
15 |
it deals with. |
16 |
|
17 |
Most devices fall into one of two groups: storage and I/O. Auto-mounters do |
18 |
not care about your keyboard, whereas X needs to know about your monitor, |
19 |
card, keyboard, mouse. Why does hal try and abstract both? Seems silly to me. |
20 |
|
21 |
One could also argue that the developer's state of mind is reflected in the |
22 |
chosen method of configuration - xml files. This just defies all |
23 |
understanding. Apart from the fact that real-world xml is almost unreadable, |
24 |
the conditions that make xml useful are simply not present in hal... |
25 |
|
26 |
xml works well when you have system A talking to system B and neither A nor B |
27 |
(nor user C) know in advance exactly what the other is. They might not even |
28 |
know much about the data schema being used, so that metadata is in the xml. |
29 |
This is so completely not the case with hal on a local machine, that it defies |
30 |
description why the dev thought it might be useful. |
31 |
|
32 |
-- |
33 |
alan dot mckinnon at gmail dot com |