[ge-talk] Get login details while booting.
Truls Becken
truls.becken at gmail.com
Fri May 11 03:41:47 EDT 2007
On 5/11/07, Mat Hounsell <mat_geek at yahoo.com.au> wrote:
>
> It takes a few seconds to type in the login & the password. In that time
> it should be able to bring up the boot fs and the network stack to do the
> authetication.
>
> Currently you have to wait for everything including the identifying
> server.
>
> Display a login box and password box on the boot screen above the progress
> animation.
>
>From the user's point of view, this idea seems like a natural way to improve
user friendliness. A system designer's view, on the other hand, would
probably be that this is both dangerous and somewhat hard to achieve. This
is because what you said implies making the login interface part of the
kernel, while traditionally it's in user space - at least on Unix-like
systems.
I think the key here is to make the part in the kernel do as little work as
possible. If it was somehow possible to take a user name and password at the
boot screen, and inject that into the login panel later, it could perhaps be
feasible. Maybe one way to do this is to keep the data in the input buffer
where it would be read when the login panel was displayed? The input server
would have to be tricked into waiting for the correct text field to have
focus before starting to read the input buffer, though, so this hack is
probably not a suitable approach at all.
There are questions to be asked about such a scheme as well, like:
1) What are the security concerns about transferring credentials read at
boot time to the login application?
2) What if you mistyped something at the boot screen? You will not know
until the login panel comes up and tells you to try again.
3) What if you started typing late in the boot process and only got halfway
before the screen went blank to start the real interface?
The time you have to wait for the login panel varies alot between operating
systems. Many do all their startup actions, including waiting for response
from a DHCP server, before you're allowed to login. Others actually try to
postpone as much as possible. Windows XP isn't too bad in this regard,
although it doesn't seem to do much while waiting for user input.
With BeOS traditionally being fast to boot, maybe the best approach simply
is to get the login panel up as fast as possible or even faster *, and then
continue starting up services while the user is logging in?
It would be cool, and also quite innovative, to present a login prompt at
the boot screen, like Mat suggests, but I'm so sure it's feasible. The
questions above, and probably others I didn't think of, would have to be
answered first. What do everybody think?
*) meaning it's OK to cheat as long as you don't break anything / get
caught.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.bug-br.org.br/pipermail/glasselevator-talk/attachments/20070511/cfb454ff/attachment.html
More information about the glasselevator-talk
mailing list