Conditions and symptoms are:

• You are using Network Manager (e.g. under Gnome 3)
• Your network connection shows as “connected” in Gnome “System Settings”, “Network” and displays an IP address, Default Route and one, or more DNS servers.
• You can successfully ping external addresses in a terminal session:
  e.g. ping –c 3 8.8.8.8
• DNS lookups fail:
  e.g. nslookup www.google.com.
• You may see an SELinux AVC denial alert under Gnome.
• The file /etc/resolv.conf.tmp exists and contains a valid DNS server IP within.
• /var/log/messages contains the following:
• <warn> could not commit DNS changes … Permission denied
• setroubleshoot: SELinux is preventing NetworkManager from unlink access on the file /etc/resolv.conf

To fix this:

Open a Terminal window session and type:

su –
<enter your password>
restorecon /etc/resolv.conf

Now disable and re-enable your network connection from the Gnome network tray icon and re-test DNS lookups by either opening up a browser and navigating to any page, or by using nslookup in a terminal window.

The detach method doesn’t seem to work for me. I did find that eject -s sr0 (or 1,2, etc. depending on what has been assigned to the fake CDROM “ConnMgr Storage”) does switch the device, though – but then the device has had a firmware update, even though the hardware is an earlier version than what is on sale today.

>lsusb | grep D-Link
Bus 001 Device 018: ID 07d1:a800 D-Link System

>eject -s sr1

>lsusb | grep D-Link
Bus 001 Device 019: ID 07d1:3e02 D-Link System

This is under Fedora 14, 64 bit .

A little bit more digging and I’ve come up with:

########################################################
# D-Link DWM-156 HSUPA 3.75G USB Modem
# 07d1:a800 07d1:3e02
# H/W Ver. A1
# F/W Ver. 1.04 Upgraded to 1.05b01
# Upgrades at ftp://ftp.dlink.co.uk/dwm_3G/DWM-156/
# 27-Jan-2011

DefaultVendor= 0x07d1
DefaultProduct= 0xa800
TargetVendor= 0x07d1
TargetProduct= 0x3e02

DetachStorageOnly=0

# Standard SCSI eject
MessageContent=”5553424312345678000000000000061b000000020000000000000000000000″
MessageEndpoint=0×01
ResponseEndpoint=0×81
Interface=0×00
NeedResponse=0
CheckSuccess=20

Normally, tightening the truss rod (clockwise, viewed from top for a top adjusting truss rod) will reduce the gap between string and fretboard (the neck arcs back). Conversely, loosening the neck (anti clockwise) will increase the gap between string and fretboard (the neck arcs forward). You’ll need an allen key to make this adjustment.

Put a capo on the first fret, then finger each string on the highest fret and check that it barely touches the frets in between the capo and your finger. If you play the upper part of each string (whilst still fingering the highest fret and with the capo still on the 1st fret) it should be playable with only a slight buzz. You may find that the lowest string (6) and the highest string (1) have different clearances to their respective frets. If this is the case, it might indicate a twisted (or otherwise distorted/warped) neck, which is difficult to fix at best. When checking the neck it is best to have new, or fairly new, strings fitted and always recheck clearances and tuning between adjustments.

Don’t forget to do a final re-tune and check, before moving on. Finally, re-tighten the back-plate screws to the same original tension (should be about a turn, if that’s how much you loosened them in the first place). Once you have set-up your neck, then you can go ahead and proceed with any bridge adjustments.

So I decided to open it up and inspect it for obvious faults. Everything seemed in order, nothing touching anything it shouldn’t and relatively well-built considering the price I paid. The overall layout looked sensible. Fused input, MOV protected, decent connections and reasonable PCB tracks, big caps, chunky coils, a myriad of power MOSFETs/transistors and a couple of OPTO-isolators. Some surface mount technology on the underside of the board, completing the picture.

The only thing I can put this leakage current down to, following my cursory inspection, is that somewhere they have used a capacitor, or two, between the Live/Neutral and internal ground connections to filter out noise (switch-mode PSUs create a lot of it).

Since it is a class 2 PSU, it has no earth connection. Just a two-pin plug which has no effect on the phenomena whether you plug it one way, or another.

A quick Google only turned up this thread, on the IET board which briefly discusses the issue, but seems no more than a “shrug, oh well”.

Perhaps this is brought about by the fact that some existing standards seem to permit this kind of design. Depending on the standard, as much as 1.5mA RMS at 277 VAC at 60Hz (EN/IEC) is allowed. Some standards are tighter, allowing only 0.25 mA.

I don’t know about anyone else, but 1.5mA AC at mains voltage leaves me a little uncomfortable, not least because it’s not always dry un-broken skin (at several KΩ) that could become exposed to it, but bare feet, a sweaty hand and a fresh cut could present as little as a couple of hundred ohms resistance to ground via vital organs.

I decided to see how bad the problem was. First, I checked the resistance between each mains input pin and the PSU outputs. My digital multi-meter (Uni-T UT71D) was unable to register any reading – meaning that the isolation DC resistance had to be well over 40 MΩ. Good enough for me. So far, so good. Next up, I measured the voltage from the output ground to mains ground (via the earth tag on my ‘scope). this came in at a whopping 93 VAC RMS and 93.1 VAC+DC RMS. Ok, ok, no need to panic until we’ve checked the current on a near-dead short to earth – unfortunate choice of words not-withstanding. I ran through the current scales on my meter, amps, milliamps and finally micro-amps, where I eventually got a 0.35 µA reading.

Given the very low current reading I figured I could just about live with that, even though I don’t like it one bit, no sir! I’m left worrying about the consequences should those caps seriously degrade, or ever fail (and they do from time to time).

Also, I can’t quite reconcile the lab measurements to the fact that I can detect this leakage current with the tips of my fingers. Perhaps it’s those whopping 90 odd VAC – which, by the way, are enough to light-up the neon light of an electrical screwdriver (you know the old-fashioned kind that you have your finger on the end of and have a resistor in series between you and the neon).

My conclusion is that this is the way it is and you won’t know until after you have bought product x, since this seems to be common practice and acceptable under current standards. I’m still not happy and I may decide to add an earth wire to the casing. The leakage will still be there, but at least it’ll drain away to ground and if those caps ever fail, whoever is using this thing won’t have such a nasty surprise.

As an aside, I tested a few smaller power adapters (the ubiquitous wall wart) and found them much tamer, with voltages of only around 10VAC bleeding into the output.

Meanwhile, ISPs seem to be dragging their feet in announcing any sort of IPv6 service to end-users. At least that is certainly the case here in Spain and from what I can see, throughout Europe.

Currently, the large service providers, such as Google and Facebook, offer us only IPv6 modified URLs (e.g. ipv6.google.com and www.v6.facebook.com) to access their pages. They are in dread of adding IPv6 DNS records to their IPv4 domain names in case they break, or add delays to, their user’s access to their content. Others, such as Yahoo, have proposed only giving IPv6 addresses over IPv6, leaving IPv6 queries over IPv4 out in the cold.

The IPv4 pool has now been exhausted. IPv4 addresses will only be had at a premium. The time for IPv6 is now overdue. What can a simple mortal do against such indifference, apathy, fear, uncertainty and doubt?

One answer is to stay off the net on IPv6 day. Go out and enjoy the summer sunshine. That way you won’t have to suffer the headaches of whatever goes wrong when major service providers turn on IPv6, for their primary Internet domain presence. And don’t think that you can weather the 24-hour tempest by using just IPv4. It won’t necessarily keep you safe. If you follow this course of action, you can return on the day after happy in the knowledge that, after IPv6 day, the luddites will switch back to IPv4-only.

However, if you are brave and possibly as foolhardy, you can revisit your IPv6 connectivity and join in the fun. There appears to at least one tunnel broker that may yet provide some relief to the unfortunate, namely HE.net (Hurricane.Electric Internet Services). I shall be trying out their services in due course and I will report back here.