[systemd-devel] Network Interface Names: solution for a desktop OS

[Michael Biebl]

So why don't you implement such a scheme? Talk is cheap

2016-04-10 18:22 GMT+02:00 Xen <list at xenhideout.nl>:
> I just want to present my conclusion here succintly.
>
> https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/
>
> Was introduced to safeguard against a rare occasion where an important
> network computer in a critical environment would suffer from the kernel
> anomaly that assigned network interface names may be unreliable.
>
> In any system not configured by the system administrator to cater to
> this issue, even simply forgetting to do it could result in some
> security breach.
>
> This is how I read the issue today. The designers wanted to create
> something that would never fail, and never cause this scenario to happen.
>
> But now all desktop users pay the price of having incomprehensible names
> for their network devices, and all server administrators who like
> writing scripts or defining firewalls have to deal with names that are
> going to change from system to system.
>
> There is a solution to map the original hardware devices (not the new
> names) to more meaningful names by manual choice. However, this is not
> usually implemented and I do not know of any distribution that does this
> by default.
>
> The solution requires the user to either know about the way to find help
> (man systemd.link) or to look it up on the web (and know that systemd
> causes it). Then he or she needs to create multiple files for each
> interface containing either the hardware address (MAC) that is easy to
> find, or the PCI bus address according to udev that is less easy to find.
>
> While many people would probably like a more comprehensible naming
> scheme, the burden is now on each individual user to create it, while a
> consensus on a regular scheme would not be hard to reach.
>
> For instance, calling wired internet devices "ethernet0" while wireless
> ones would be called "wireless0" would not be an odd thing to do at all.
>
> So what I am simply calling for is for distributions to make a choice in
> the names they want, and then to configure it by default.
>
> PCI hardware names can be mapped to predictable names, most likely. MAC
> addresses could equally be used.
>
> The only thing to decide upon is the actual naming scheme. As said,
> "ethernet#" and "wireless#" would be obvious.
>
> It wouldn't take more for a distribution's installer than to find these
> devices and create mapping files based on them.
>
> The configuration with multiple files in /dev/systemd/network is not
> easy, but for an installer that would not be an issue. The configuration
> is not easily discoverable by any user traversing the filesystem, nor is
> it intuitive to use multiple files for a single configuration but a
> likelihood of a user needing to change it, would be very low.
>
> If you want systemd to make sense, you must make it easier for users.
>
> The names I propose would be easy to understand and contain no risk for
> the scenario I described unless hardware is removed from a system (but
> not put back).
>
> And it might not even have that issue.
>
> Comprehensibility increases legibility and it could even reduce the risk
> of some administrator making mistakes.
>
> What it would create on a desktop OS is user-friendly names while having
> scarcely any implication, or not at all, for the risk situation
> described. PCI bus addresses could be used by default, or MAC addresses
> if that was an issue. The kernel-based reordering as described would
> never happen.
>
> And all it requires is for the current system to create a default
> mapping that requires the installer of the system to do some work.
>
> And not much.
>
> So what I vouch for is a default mapping, that is all.
>
> A default mapping to names such as "wireless0" "ethernet0". People could
> also think about renaming "lo" to "loopback".
> _______________________________________________
> systemd-devel mailing list
> systemd-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/systemd-devel



-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?