Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Doesn't nmtui come installed by default?


Only if your distro uses network manager

This wraps iwd tools I think. Even though, to be fair to the spirit of your comment, I don't see anything in the readme offering significant more functionality than iwctl


The chances that your distro ships with network manager are multiple orders of magnitude higher than it shipping with iwd though… I don't think iwd is the default anywhere?

FWIW there's also wpa_cli which tends to work almost always (since everything wraps wpa_supplicant in the end), but it's a CLI rather than a TUI, and not a particularly usable one.


> multiple orders of magnitude higher than it shipping with iwd though

Even if that were true (and ignoring the fact that they aren't alternatives to each other, like other comments point out)

If you're on a distro with iwd you might still want a TUI without having NetworkManager installed. A tool doesn't need to be universally needed to be useful


iwd does not wrap wpa_supplicant, it's a from-scratch implementation and a much nicer one at that.


It looks like it wraps NetworkManager or ConnMann - both of which wrap around wpa_supplicant for wifi. So yep, iwd wraps wpa_supplicant.

Architecture diagram on home page: https://iwd.wiki.kernel.org/


You're reading that diagram backwards.

NetworkManager and ConnMan can optionally _USE_ iwd as a backend _INSTEAD OF_ wpa_supplicant.

iwd does _NOT_ use NetworkManager/ConnMan

Source: gentoo user who explicitly avoids the buggy disaster that is NetworkManager+wpa_supplicant whenever possible


I noticed my error. To be fair, I don't think I read the diagram backwards. I think the diagram is drawn backwards instead. In it, iwd seems to talk to the iwd backend via D-Bus as they're close together.

Or maybe that's a diagram technique I'm just not used to.


It reads as a fairly normal diagram to me; NetworkManager has an iwd backend component/plugin that talks over D-Bus to iwd, which in turn stands on top of ell which in turn stands on the kernel (which itself contains a bunch of components of interest).


NetworkManager can use iwd as a wifi backend instead of wpa_supplicant, but nm isn't needed as iwd can also manage the networks on its own. iwd should never run at the same time (on a single network interface) as wpa_supplicant, as wpa_supplicant is (almost?) entirely superseeded by it.


Oh ok, so I misread the arch diagram. I thought that iwd was talking to the iwd backend, which would then delegate to NM or CM.

It looks like you're right [0], so I stand corrected.

It also looks like it can't fully replace wpa_supplicant though:

> IWD and the NM backend are work in progress and the capabilities are still limited.

[0] https://iwd.wiki.kernel.org/networkmanager


That paragraph of the article looks to be ~6 years out of date according to the network manager version number it lists, around the time of the initial iwd release, and the whole article seems to be at least 2-3 years out of date since then iwd is well into version 2.x now.

Distros like Ubuntu have defaulted to iwd as the NM backend for Wi-Fi for a couple years (and now in the LTS version). It really is a quite popular and stable replacement to wpa_supplicant.


Alright, thanks for the information! :)


  > everything wraps wpa_supplicant in the end
What should I be reading to understand exactly how the pieces fit together?

I'm on Kubuntu, and have been suffering wifi disconnects for years across multiple computers with multiple USB wifi dongles and multiple access points. Literally not a bit of hardware shared, not even the access points. I suspect that Kubuntu shuts down USB wifi dongles after a period of time, as the only reliable way to get back online is to simply disconnect then reconnect the USB dongle. I forget what dmesg says (on mobile now) but it's something to the effect of a voluntary disconnect.


> wifi disconnects

This is more likely to be a driver issue than management issue.

I think, this was posted to HN before, but I cannot find the article right now, but the gist of it is: both WiFi h/w vendors and Linux driver authors are to blame for the dismal state of events. On the h/w vendor side everything as per usual: no drivers, or broken drivers, incorrect product labeling or self-identification, false or imprecise advertisement of available features. On Linux side: slow inclusion of community-provided drivers for various WiFi modems, no real effort at cataloging and unification / systematization of available drivers.

My experience so far is this: Intel h/w is the one that has best Linux support. So, if you are on Realtec or Mediatec, you are getting the short end of the stick.

Another universal problem: WiFi protocol keeps iterating at a relatively high speed. Which means that you are often in a situation where either your modem or your router don't speak the same protocol at a good level, and the middle-ground that they find is not well-supported or, sometimes will only be found after lengthy unsuccessful attempts by both sides to do the handshake. This is often also exacerbated by the management software which may misidentify these "handshake problems" and "solve" them by restarting something in the network stack, essentially sending you into an infinite loop of trying to connect.

In my unfortunate case with my latest PC build, I had to swap three WiFi modems before I found one that worked satisfactory with the router I have. So, if you have some extra $ to throw at your problem: maybe try getting another WiFi modem?


Thanks, I'll try to find a good USB or PCI Intel wifi modem. Good idea.


We had disconnects in our office (so very different machines, different Linux distros and a Mac at least). Don't remember for sure, I believe the message was reauthentication successful and after that connection was lost.

The AP runs OpenWrt. We worked around it by changing the reauthentication interval to 8 hours. Those working longer have to deal with it... Yeah, it's probably a weaker security now...


I've never had that problem on Kubuntu, so it's not a general Kubuntu issue. Multiple different access points, but just one PCIe wireless adapter (identified as "Qualcomm Atheros AR93xx Wireless Network Adapter").


Is that adapter on the motherboard, PCI, or Usb? I'm starting to suspect that the problem is with the USB connection.


It's a PCIe card. Bought in 2016, seems to be this: https://www.tp-link.com/us/home-networking/pci-adapter/tl-wd...


Perfect, thank you.


W Linux Supported Hardware

W Linux Solving problems


iwd is an alternative to wpa_supplicant, not NetworkManager.

NM has support for iwd.


> The chances that your distro ships with network manager are multiple orders of magnitude higher than it shipping with iwd though… I don't think iwd is the default anywhere?

It doesn't matter what the default is, though. Some people don't use NM, and this tool is for them.

> (since everything wraps wpa_supplicant in the end)

That's not true. IWD doesn't require wpa_supplicant.


Well, I showed off my ignorance of iwd quite effectively :)

TIL it is a wpa_supplicant replacement. Sorry for the misinformed comment!


wpa_supplicant is a buggy hot mess with very little effort in the way of accepting patches, from me or anyone else.

Iwd is a separate program and actually works.

Try it out. It doesn't suck


Yes, but nmtui is terrible and I for one would welcome a replacement


What do you feel is terrible about it? It has always done exactly what I needed it to do, while being much easier for occasional use than nmcli




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: