Category: Uncategorized

Any idea how much bandwidth a Youtube stream requires? Me neither so I recently took some time to have a closer look. Before reading on, take a guess.

Obviously, Youtube does not stream videos in the low compression format in which they are usually uploaded. In a test I did, the 1280x720p video I took with my smartphone had a data rate of about 11 MBit/s and a file size of around 45 megabytes for 32 seconds. Once uploaded and converted by Youtube the HD version of the video (720p) streams at around 2.7 MBit/s and has a total file size of around 11 megabytes. That's the format used by high resolution smartphones today. The default format (480p) version of my test video streams at around 1.2 MBit/s. I've also compared it to other professional content such as movie trailers and the data rates are about the same.

Interesting numbers and the next higher resolution is already in sight, 1280p. As for know, however, that is still rather the exception.

Lightreading recently reported that US cable network operator Cox has given up its plans for wireless after having bought LTE spectrum for $550 million previously and launching as an MVNO on Sprint's network about a year ago. That was a short and costly trip to wireless land.

It seems that it becomes harder and harder to get established as an additional wireless network operator in the competitive landscape today where your competition already has installed nationwide networks and a solid subscriber base. Other examples I can think of with high plans and quick folds or non-delivery afterwards were a number of companies in Germany who bought WIMAX spectrum a couple of years ago but never made it to a real roll-out. I guess similar things happened in other countries a few years back as well.

A current new startup that I keep watching closely is Free in France, a DSL network operator currently trying to establish itself as the fourth network operator in a not so competitive wireless market. But building a network that customers actually accept or partnering with another incumbent until the network is built out sufficiently are difficult tasks. I wish them all the luck because France could use some wireless competition from the customer's point of view. And after all, every rule needs an exception!

Rather than new network operators appearing on the scene the opposite has been happening in recent years with some incumbents either merging or getting permission to do some form of network sharing beyond just using the same towers. For both regulators and network operators this is a tricky thing to get right because once networks are merged its difficult to impossible to reverse. Also, once networks are merged but companies are still separate, strategies for network evolution are closely linked, perhaps too closely to differentiate or to do what a company really likes to do. Let's hope such moves do not fire back for the regulator, for the companies involved and least of all, for the users. In wireless, four (real networks) seems to be the lucky number, not three, when it comes to a balance of good prices, service and coverage on the one side and making a profit on the other.

Austria has long been the country with one of the lowest prices in mobile in the western hemisphere and give you a pretty good idea of how things might look like in other countries in a couple of years from now. The latest offer there from one network operator is 10 GB of data per month for 9 euros on a prepaid SIM card (with automatic monthly renewal). Data rates are throttled to 4 MBit/s in the downlink direction and 2 MBit/s in the uplink direction but that is more than enough for most things. And on top they are explicitly mentioning the possibility to use the SIM in a 3G dongle bought elsewhere. Makes you think when you see tariffs and the terms and conditions in other countries…

One of the good things of most Linux based distributions is the central software catalog and combined update capability so the user doesn't have to update different programs from different resources or be bothered by programs individually to start or complete an update (like in Windows for example). While this is ideal and convenient for most users, the downside is that if the distribution decides that a package is in maintenance and security update mode only, there are no feature updates. One of those programs for me is Wireshark in which, over the last one and a half years, a number of interesting decoding packages for GSM have been added which are not available over my default Ubuntu 10.04 software updater.

Some fast developing programs such as Firefox have Linux binaries for download and can thus be kept up to date by the user if he wishes so in a pretty straight forward way. Wireshark, on the other hand only comes as source code for Linux so you have to compile it yourself if you want to have the latest and greatest version. So far I always shied away from this, too much work, too much hassle, too many unknowns. But these days the "feature pressure" became too big so I used an Ubuntu in a virtual machine for a trial and error compile. Turns out it was not as difficult as I thought but not straight forward either.

After downloading the sources and putting them in a folder in my home directory I checked the web for compile instructions and found them here. I guess things are made easier if you already have a relatively recent Wireshark version installed via the standard software repository so the general procedure only consists of the following three shell commands:

./autogen.sh
./configure
make

I had to run the process a number of times because I didn't have two software components installed for the scripts. One was gtk+ where the script wanted a version 2.4 or higher. Unfortunately, Ubuntu 10.04 only has a lower version number available in the software repository. So I installed that and it worked fine as well. After that all you need is some time depending on the speed of your CPU. Once done, wireshark can be started from the directory with ./wireshark (don't forget the ./, otherwise the old version is started from the default path).

And finally, the directory that contains the Wireshark executable and can be copied to other machines running Ubuntu 10.04 so no compile process is necessary there. Excellent!

As a frequent worldwide traveler I find it quite annoying when I come to big cities (such as that hosting the next summer Olympics for example) and the tube system is lacking wireless network coverage. The issue is more prevalent in some countries than in others and now a recent study by the New Cities Foundation on where tube systems have coverage sheds some interesting details on this topic. According to their results South Korea and China have 100% network coverage in big city tube systems, followed closely by Japan with 87.5% and Asia as a whole with 84.4%. Europe is pretty much down in the list at 56% only trailed by a few such as the US where it is even worse. A result I would not have quite expected. I guess the message is clear: European and US cities beware, you are by far surpassed by your Asian competitors!

The project on Internet connection sharing over cellular networks during meetings I reported on in two previous posts is not quite over yet and I've been working on refining my setup ever since. One thing that often goes wrong in non-optimized hardware setups is the number of devices the built-in DHCP server can give out IP addresses to. In many home grown routers this is actually quite limited, even if the DHCP server range can be specified from .1 to .254. To find out exactly where the limit is, here's a script I've came up with that can be run under Linux and will send 268 requests to the DHCP server, each with a different Ethernet hardware address. Some routers stop giving out IP addresses after 50 leases, some after 60 and only few go to the full range.

#!/bin/bash

for l in {1..99}
do
  sudo ifconfig eth0 down
  sudo ifconfig eth0 hw ether 70:00:00:00:00:$l
  sudo dhclient -r
  sleep 2
  sudo ifconfig eth0 up
  sudo dhclient
  sleep 3
  echo $l " "
  ifconfig | grep "inet addr:192.168."
  ifconfig | grep "HWaddr"
done

for l in {1..99}
do
  sudo ifconfig eth0 down
  sudo ifconfig eth0 hw ether 70:00:00:00:01:$l
  sudo dhclient -r
  sleep 2
  sudo ifconfig eth0 up
  sudo dhclient
  sleep 3
  echo $l " "
  ifconfig | grep "inet addr:192.168."
  ifconfig | grep "HWaddr"
done

for l in {1..70}
do
  sudo ifconfig eth0 down
  sudo ifconfig eth0 hw ether 70:00:00:00:02:$l
  sudo dhclient -r
  sleep 2
  sudo ifconfig eth0 up
  sudo dhclient
  sleep 3
  echo $l " "
  ifconfig | grep "inet addr:192.168."
  ifconfig | grep "HWaddr"
done

Note: Only run this script on your own setup that you can reset afterwards as once the script is through the DHCP server won't give out further IP addresses. Something you don't want to have in a life network…

Here's an interesting link to a post by Troy Hunt with a detailed look on how wasteful some apps are with the amount of data they transfer, how they don't care about the privacy of the user and they transmit data unencrypted, a special treat for anyone using Firesheep in open Wi-Fi networks. Leaves me speechless. For the details head over and be frightened.

In the past months a number of mobile network operators in Europe have switched-on Wideband-AMR in their 3G networks which offers superior sound quality between two WB-AMR mobiles. On paper the difference is tremendous with the frequency range of around 3 kHz extended to 7 kHz, thus transmitting voice and music much clearer than we know today. But how does one perceive the difference in practice?

A hard environment to test this is for example in a train with background noise and recently I had a WB-AMR call while being just in such a place. In short, the experience was incredible for both parties. While on the train I had no difficulties at all to understand the other end and the voice was crystal clear. The positive impression was reinforced when the call dropped due to a coverage hole and when subsequently re-established in the GSM non WB-AMR network I had to make an effort to understand the other side and crank up the volume. I always thought narrow band AMR was quite o.k. but once you've heard wideband it suddenly doesn't sound so good anymore until after a while you ear adapts to it again. 

The title of this post is a bit controversial but I can't help it, when I see the names of companies on some devices today I wonder if my first impression of a new device wouldn't be more positive (or at least different) if I wouldn't see the company name so prominently displayed!? Let me give you some examples:

• Whenever I see an HP notebook these days, I think, "Phew, that's the company that wants to get out of the PC business, this is one of the last" (yes, I know they since have said these plans are off the table). Time perhaps heals wounds.

• Whenever I see "Philips" on a product it immediately goes back into the shelf because I will never buy a Philips product again in my life since the time they refused to give me a replacement for lost in-ear plugs for my noise cancellation headset when I had lost them due to them falling off all the time. A replacement part that perhaps costs one cent they refused to give or sell to me so I could continue using their headset completely destroyed their brand image for me permanently.

• Nokia: So sad, whenever I see Nokia these days I think, "yeah, once a great company, being the leader in the mobile domain, doing open source, being open, creating great devices, etc". But with their 180 degrees turn-around earlier in the year to the ultra-closed Windows Phone OS, the slaughter of Meego and Symbian and their subsequent ultra quick move from market leading company to insignificance makes me feel sad when I see a Nokia logo these days.

• SonyEricsson: Now that Ericsson has sold their stake to Sony and devices are likely to be re-branded soon, anyone with a SonyEricsson phone in their hand will immediately be identified as "out of date".

• HP/Palm/WebOS: Out of business, just like that…

Perhaps you think these thoughts are shallow and one should look behind the company names. But I think the opposite is the case. Companies keep changing their strategies and leave an impression to consumers (or at least to me) that no matter when I buy something from them, the company is going to change in their core values (read: not only their marketing campaign) the second I've swiped my credit card through the reader at the cash register and the device I bought and the ideas it was built on top will not get support from the company in the future and will not see any successors.

Perhaps with the mobile industry evolving at such a quick rate, such twists and turns in very short succession are unavoidable as companies search for ways not to fall behind and gain market share. As a consumer, however, this makes me feel uneasy.

For the second time in so many weeks I've been over in the UK recently and noticed that I am severely restricted in which networks I can select. Not that I can't chose between several networks but in all except one I only got voice service while data seems to have been turned off by my home network operator (read: GPRS service is not allowed). That's quite strange as I haven't seen this anymore since GPRS started many years ago and it was an issue of roaming agreements and links not having been in place.

So what could be the reason this time around? Perhaps, and it's just a guess, this has got something to do with data roaming charges having come down quite a bit in recent years in the EU and network operators who won't give each other reasonable roaming T&Cs now just disable data roaming to prevent making a loss!? After all, the theoretical upper limit in GPRS roaming wholesale prices is still a ridiculous 80 euro cents per megabyte. If you are offering something like to 200 MB roaming data for 10 or 15 Euros to your customer, roaming partners insisting on anything near the wholesale limit are clearly unacceptable.

Let's hope this is only a temporary situation as especially in the UK, with their patchy networks, it's a good thing to be able to use different networks.