Re: Diskless clients and NetworkManager

[David Sundqvist <glist dystopic org>]

Quoting Marc Herbert <Marc Herbert gmail com>:
> There seems to be something fundamentally wrong in this no-connection/
> /offline thing. Since NM can be configured to manage _not all_
> interfaces (including none at all) then why are some applications
> wrongly assuming NM is always managing the entire network
> configuration? This seems to be where the bug lies and should be
> fixed.

Perhaps. Taking a quick look at the NM project webpage, it's not  
really that farfetched: "Using the awesome power and flexibility of  
dbus and hal, NetworkManager provides facilities for other  
applications like browsers, email clients, or system services to be  
aware of the network's state and adjust their operation accordingly."

That certainly could sound fairly authoritative to a presumptive  
software developer, so I can understand why they make that assumption.

> Is there other buggy applications that people should be afraid of
> besides firefox?

Lots. And I'd expect them to increase, as an app that could give an  
authoritative statement on connectivity, as far as it's possible,  
would certainly be useful. As I see it there are basically two  
possible solutions:

1) Let NetworkManager have some form of 'read-only' interfaces or  
forced mode setting, in the cases where it can't appropriately manage  
the interface. I found some threads relating to diskless clients  
suggesting that NM could not be entirely trusted to keep its hands off  
the interface so I'd be hesitant to let it manage the interface  
providing root/swap disk access.

2) Disabuse application developers of the notion that NetworkManager  
has an accurate status. The first step of which would be to change the  
project page a bit to say "be aware of NM's perhaps not entirely  
accurate idea of the network's state"

The current workaround, unless I've missed a better alternative,  
appears to deactivate or even uninstall NetworkManager, but if it's  
moving towards being a tightly integrated piece of software it might  
be more practical to have things at least not get broken by having it  
active.

So personally I'd prefer the first alternative, but it might be hard  
to find a solution that fits NM in an architecturally nice way. I  
don't know NM well enough to judge if some kind of 'read-only' or  
'locked-on' interfaces would be conceptually sane, but that would at  
least be less ugly than options to force NM to say anything, or a fake  
NM talking dbus to convince apps the system is on line.

Cheers,
David

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.