[openbeos] Re: Singleuser vs Multiuser
- From: David Sowsy <dsowsy@xxxxxxxxxx>
- To: openbeos@xxxxxxxxxxxxx
- Date: Fri, 14 Dec 2001 10:34:15 -0500 (EST)
> It's not the fact that it breaks new ground, its that there is a
> significant subset of users who need/want the functionality.
Nope, it just sounds more like a small faction of paranoid developers.
> Even though you have a multi-user system you *don't* have to use that
> functionality. Look at RedHat 7.2, (I think thats the one) the gdm
> program allows you to set a user to always log in as, so you never have to
> bother.
Until RedHat came along the OS itself wasn't very friendly, and
didn't gain widespread appeal/popularity. They made things easier
from the get-go. Why should we make things more complicated?
>For someone else they can login everytime, having the app
> protect the data is fine, assuming the programmers wish to add the
> functionality, but I'd rather have multiple layers of protection around
> myself, I guess I'm really paranoid.
>
> Personally I'd rather not go back to the windows 9x system where anyone
> can do anything, do you really want your friends comming over and deleting
> all your files?
No, but I have a an excuse to beat them silly if they try, or a reason
to pick my friends a lot more carefully. :)
Remember, even BeOS was never touted by Be as a primary OS for
anyone. It was a "companion OS", to peacefully coexist along side other
fascist and communist OSes. So if you really want to be paranoid about
your data, you can do it on another OS.
> The ability to have different users is also quite handy
> in places where multiple people use the system, each one can't see the
> other peoples work. Not that everyone is using the system at one time,
> but they all have private home directories.
OK, reality check here: How many non-dev BeOS workplaces do you know of?
How likely is that to happen any time soon? Linux is still having a tough
time penetrating businesses as a non-server OS.
> It might even be nice to be able to say, user foo may not do the
> clone_area call.
*Sigh*. So you want developers on a multi-user system to jump through more
hoops to get things done? Bleh. No, if they are competent enough to
program, give them control over the whole system. The OS is a program
to be run on a piece of hardware for the user. The OS isn't a protective
shield to fend away predators.
> I don't know how feasbile that is, but for something
> like email attachments if you could setup a user that can only see his
> home directory, dosen't have access to any potentially harm full system
> calls etc, and then tell your mail client to execute attachments when you
> want as user foo would be really handy.
BeOS is not Unix. Why should OpenBeOS be? Oh no..."harmful system calls".
Any more harmful than causing race conditions on a semaphore, or
malloc'ing in an infinite loop. BeOS was designed as a) multi-cpu b) media
OS from the get go. Care to give a cost analysis per-CPU cycle (and
memory) of the additional overhead that multi-user support incurs? I think
Be cut corners on frills like multi-user to gain speed.
> I don't think its an R1 thing, but should be considered now to make the
> later integration easier.
>
> Just my opinion,
> dan
>
>
Other related posts:
