Pwireless2 (Version 1.0)

[bill]

No jemimah,I don't hide anything.and after 3rdscr it asks me too enter SSID.Sad but true cheers

[jemimah]

Well, you've baffled me. Some network drivers are known to have problems with wpa_supplicant/Pwireless2. If it's still locking up your machine - you should probably try the network wizard instead.

If that doesn't work, it's only mildly painful to write your own shell script to bring up networking. There's a thread about configuring wifi from the command line that might be worth your time to check out.

[jemimah]

Version 2.8 has been posted.

[prehistoric]

Now running Pwireless2 beta 2.8 on a mini-ITX box with an external USB wireless adapter based on the RTL8187.

Appearance: The system I just tried this on is actually boxpup 431, which leads to a problem I wouldn't see elsewhere. The message at the top of the screen becomes part of the background for the wbar, and sticks around. There are other examples of boxpup getting confused about backgrounds, so there is an openbox bug in boxpup. Still, getting the message mixed up with active buttons isn't good. I don't feel like changing wallpaper every time I connect to get rid of messages. Could we move the announcement down by about one row of icons (40 pixels)?

Real bug: At the moment, this mini-ITX box has no wired connection; (if it were a laptop, that would be standard.) I still get messages about requesting an IP address on eth0, and another message about a possible incorrect password when there is a timeout on eth0. If there is nothing in the socket, the interface should fail to ever see carrier on Ethernet. This is like a failure at the "found a live network" step in the old network wizard. Without any signs of life, it should never attempt to use dhcp to get an address. There is no reason to expect it to use encryption, or expect the interface to want encryption, and no need for a timeout.

Next problem: when I tried to get a screen shot of the offending behavior at the top, I deliberately disconnected from my standard wireless connection. The profile for the network now exists, but it is not in use. When I press connect, the tool-tip message says the button will connect and add a profile, (which should not be necessary.) In any event, nothing seems to happen when I press it. How should I temporarily disconnect and reconnect?

[prehistoric]

...It has been many
moons since I did any C or prehistoric GWBASIC (Yes I am way past "Sale Date) and if this is in error,just blame old age. cheers

What?! GWBASIC isn't prehistoric! Even original BASIC came long after I was used to punching comment cards. (See, I did catch the clue you left when you talked about remarks on code, earlier.) You aren't even close to prehistoric. I'll bet you don't even remember when germanium diodes were pretty cool, (because they don't need filament heaters.) And, did you ever hear of recirculating memory?

Watch yourself, young whippersnapper!

[01micko]

Hi prehistoric.

I have seen that exact problem with the yaf-splash becoming part of the wbar background in Dpup, it is a problem of wbar, same will happen if you have a window open before wbar loads and it is in the area. One work around could be to increase the margin, it is in the wbar command which in boxpup I think is executed from .xinitrc.

Cheers

[jemimah]

Prehistoric, I had the same problem with wbar when building BoxPuppeee. I fixed it manually for that, but I guess we need an easier way. I'll add it to my todo list.

Would you run iwconfig and confirm that eth0 isn't there? If it shows up on the dhcp tab, you can check the 'ignore' checkbox and that will make the messages go away.

Now about the 'Connect' button... The 'Roam' button now does what the connect button used to do, since it was confusing to users if pressing 'connect' didn't actually connect you to the selected network. Also clicking the tray icon will connect/disconnect as you expect.

[gposil]

Hi prehistoric.

I have seen that exact problem with the yaf-splash becoming part of the wbar background in Dpup, it is a problem of wbar, same will happen if you have a window open before wbar loads and it is in the area. One work around could be to increase the margin, it is in the wbar command which in boxpup I think is executed from .xinitrc.

Cheers

Just right-click wbar and it will refresh it's background...

[prehistoric]

...One work around could be to increase the margin, it is in the wbar command which in boxpup I think is executed from .xinitrc.

I can find the wbar command in .xinitrc, but don't see anything like "margin". I'm also unable to find a man entry for wbar on-line, so I'm working with what I can glean from "wbar -help". What I was suggesting for jemimah was to move the message a small distance down, if possible, so it would never interfere with any bar or launcher at the top of the screen.

The manual fix, right-clicking on wbar, is a classic nerd solution. You can trace this back to a message from the developer about an undocumented feature. One suggested solution has been to automatically do this every 5 seconds. This wouldn't consume excessive resources, but would probably have side-effects. We might do something every time there is an event which might uncover wbar, but this doesn't appeal to me.

One curious aspect is that later messages about getting an IP address do not stay up there. This seems to be confined to that message about restart. If we could convince wbar this message is above it, that part of the problem would go away. This might be as simple as delaying the command until wbar gets started, and takes that first background snapshot.

Jemimah,

iwconfig shows that the eth0 interface is present, but has no wireless extensions. Having the interface does not mean any cable is physically connected.

One surprise from the output for wlan0 is that it correctly reports the SSID and a hexadecimal value for the key, but then says security is open. This network is definitely secured with WPA. Without that key, you can't connect.

I think some confusion about the connect button might be resolved with pop-up messages saying "already connected", or "reconnect using existing profile". The message about adding a profile should only appear when you connect to a network without an existing profile.

We do need a natural way to update a profile if the key changes. This takes us back to the issue of not getting a negative acknowledge when the key is wrong. Perhaps a flag which indicates whether or not the last press of connect was followed by a success could be used to ask for a new key on the second press. A message like "use existing key?" could allow users to avoid retyping when the failure to connect using an existing profile was not due to mismatching security keys. If the user does retype the key, that should update the profile.

"Roaming" is a word with vague connotations currently being exploited by cellphone companies to double advertised rates, or halve purchased "minutes". (Are cellphone "minutes" a new, and unregulated, currency?) It is possible to be hit with roaming charges at your home location, depending on which tower was used for original setup. Because these tricks wouldn't work if people understood precisely what it means, I'm afraid no commonly accepted meaning exists, making that button less informative than intended.

[jemimah]

I think wbar is taking a screenshot to use as its background. In that case, adding a 'killall yaf-splash' to the startup script right before wbar should fix it.

Prehistoric, apparently your kernel is telling dhcpcd that eth0 is present and has a carrier. Pwireless2 doesn't have any code that would do this.

Apparently 'open' security means different things for different cards. From the iwconfig man page:

The security mode may be open or restricted, and its meaning depends on the card used. With most cards, in open mode no authentication is used and the card may also accept non-encrypted sessions, whereas in restricted mode only encrypted sessions are accepted and the card will use authentication if available.

The connect button should not ask you for a password if one is already stored. What it does offer you is a way to manually specify which network you want if multiple ones are in range that have profiles. I will update the tooltip to explain better.

I did add code that will tell you if wpa_supplicant failed due to a bad password. It only checks when you are adding a profile. It assumes you will know what to do your password changes. If you WEP key is wrong, yaf-splash is supposed to tell you.

What should I use instead of 'Roam'? 'Auto' maybe.... The tooltip does explain what it means.

[prehistoric]

Jemimah,

While I was fooling around with wbar, I changed the .wbar configuration file. When I restarted X to test the changes, the network connection stayed up, but the pwireless2 tray icon showed it disconnected. Clicking on the tray icon changed the icon, and left the connection working. However, some seconds later I got a pop-up message about a possible bad key. Seems restarting X confuses the poor thing about the status of network interfaces.

Prehistoric, apparently your kernel is telling dhcpcd that eth0 is present and has a carrier. Pwireless2 doesn't have any code that would do this.

Even so, where does it get the idea a wired connection requires a password? iwconfig reported no wireless extensions.

Still thinking about "roaming". If you select the network before pressing "connect", that takes care of one meaning. If you are telling it the profiles may not be the same, even if SSIDs match, (a common situation when people leave the default SSID alone,) shouldn't that force it to create a new profile? You then need to select profiles based on something more than SSID. Where ambiguity exists, you need to ask whether to use an existing profile, or create a new one.

[prehistoric]

I think wbar is taking a screenshot to use as its background. In that case, adding a 'killall yaf-splash' to the startup script right before wbar should fix it.
...

This solves the immediate problem, but I would bet there will be other instances. Shouldn't yaf-splash have a hook for a restore action when the message times out? In many, cases this would not be needed because the window manager would handle things correctly. In case a special action is necessary, attaching a command, (equivalent to a manual right click on wbar in this case,) would be a neat solution.

[jemimah]

Even so, where does it get the idea a wired connection requires a password? iwconfig reported no wireless extensions.

Yep that's a bug. It's only supposed to say that for failed WEP connections.

Still thinking about "roaming". If you select the network before pressing "connect", that takes care of one meaning. If you are telling it the profiles may not be the same, even if SSIDs match, (a common situation when people leave the default SSID alone,) shouldn't that force it to create a new profile? You then need to select profiles based on something more than SSID. Where ambiguity exists, you need to ask whether to use an existing profile, or create a new one.

Hmmm, that's asking a lot, because you can't use BSSID to differentiate either, since the same network might have multiple APs. How do other network manager's handle this? Currently if this happens, you need to modify or delete the existing profile.

[jemimah]

When I restarted X to test the changes, the network connection stayed up, but the pwireless2 tray icon showed it disconnected. Clicking on the tray icon changed the icon, and left the connection working. However, some seconds later I got a pop-up message about a possible bad key. Seems restarting X confuses the poor thing about the status of network interfaces.

Yeah, a real programmer would have written a proper applet. Gtrayicon is really a duct tape and bailing wire solution, one that definitely leaves something to be desired. I've been getting some exposure to Vala lately. Maybe I'll add a real applet to my todo list too.

Shouldn't yaf-splash have a hook for a restore action when the message times out?

I don't know what you mean by this.

[01micko]

I can find the wbar command in .xinitrc, but don't see anything like "margin".

Sorry prehistoric, my mistake, the "offset" option is the one you want.

Perhaps a "killall yaf-splash&& wbar -[options] " might do it.

Cheers

[prehistoric]

...
Hmmm, that's asking a lot, because you can't use BSSID to differentiate either, since the same network might have multiple APs.

Good point.

How do other network manager's handle this?
Currently if this happens, you need to modify or delete the existing profile.

If you're asking about a M$ network manager, I have no idea what they're doing much of the time.

I haven't gone looking for other open-source network managers, because this is an opportunity to rethink the whole business instead of copying code, and watching the size grow. Now that we have a very specific question, I'll see what I can find.

Using nothing more than my own brain and experience, it would seem the combination of SSID and key identifies the network. Multiply APs have different BSSIDs, as you point out. If someone changes the network channel, because of interference, but leaves SSID and key the same, it should be treated as the same network.

If you are roaming, and encounter another network with, e.g. SSID "linksys", and no key, it doesn't matter if you use the same profile. If it has a key which does not match a previous network, that calls for a new profile. This would give us "linksys (0)", "linksys (1)", etc. (Anyone remember how Panera hot spots work?) Once multiple profiles exist, it is easy for a user to select one. The catch here is knowing whether to create a new one or modify an existing profile. Until you try to connect, and fail, you don't know the key is different. This takes us back to the lack of a negative acknowledge. The problem may be more trouble than a solution is worth.

[jemimah]

I'll have to test wpa_supplicant can handle that. If it will, I'll modify Pwireless2 to do it that way.

[prehistoric]

After watching someone misinterpret the tray icon, I'm adding a request for three-state indications: connected, trying to connect, disabled. People will just sit there staring at the icon, (and wondering what is going on,) for minutes before they figure out that it isn't even trying because they clicked on the icon, and thus explicitly told it to disconnect. When they want it to connect, they need feedback that this is in the progress, or they will take further action, possibly even forcing it to disconnect.

It goes without saying you and I never do anything so stupid. Like many successful politicians, I'm asking this for all the little people out there. (In the Puppy Leprechaun group?)

[jemimah]

Unfortunately, Gtrayicon only supports 2 states. It might be less confusing not to have an applet at all, but it is convenient once you know what the deal is.

[jemimah]

Beta 2.9 has been released. This should put an end to dropped connections. You will still see green messages about the drops, but the connection won't actually drop now. The messages will be removed in the next version but I have left them in for debugging purposes in this version.

The applet will now work correctly if you restart X and the connection isn't lost.

Some WPA networks that didn't work before may work now.

This is a major philosophical change for Pwireless2 so please let me know if you have issues.