DragonFly kernel List (threaded) for 2004-03
Re: Unionfs etc Re: Packaging
I was speaking from a more theoretical level and not so much practical
in relation to what we currently use.
I mean for example -- OS's on much older computers -- lets say the
Commodore 64 (maybe not the best example) -- the user did not know the
OS was running. They interacted with it without having to know about
any of its underlying structure. The point is that no unnecessary
complexity is around to make simple tasks confusing.
This is more of a user-interface issue. For example, you don't need to
know how your car engine works in order to drive a car. This is
abstracted by the car interface -- the key, steering wheel, gas pedal, etc.
Currently many *nix OS's require the user to know much more low-level
stuff than should be necessary -- even for server use.
My idea was more to seperate system-software from user-software, so that
the more obscure and unnecessary complexity could be ignored.
Maybe I've been reading too much Raskin.
Joerg Sonnenberger wrote:
No. This is wrong. You can use views to restrict users to view only
programs they want to use, the needed shared libraries and some devices.
But you can't enforce an even stricter view (e.g. only the documents) since
the applications just won't work anymore.
Such a strict "filter" belongs into userland applications and only there.
I'm sure these ideas could be implimented easily enough through
metadata, extended attributes, or whatever you want to call them.
We have quite a few things [or will have] which could make good use
of extended attributes. Nobody volunteered yet to part it :)
Some of the things BEOS did with metadata were nice.
I'd like to use EAs for security related features, but there quite a few
other uses as well.