In my living room I have a big flat screen TV connected to Linux MythTV server, which I use for recording and watching TV and DVDs.

Thanks to the long weekend I had some time to attach a webcam (VF0415 Live! Cam Vid. IM Ultra) to that computer, mount it over the TV, and make it work (nothing to do, really) with Skype.

It works great, but I still needed to get out of MythTV and use mouse/keyboard to access Skype and make calls. I am sure it would not be that hard to create a proper plugin to make MythTV work with the client-side Skype API, but it probably makes little sense now that they are about to release their SkypeKit platform and I am sure someone will just convert the good SIP plugin for that.

Here is how you can add an entry in the Main Menu of MythTV to start Skype:

• Find and save somewhere on your disk a reasonably sized Skype logo in PNG format
• Add to “/usr/share/mythtv/themes/<your_theme>/menu-ui.xml” an entry like the following

  <state name="SKYPE">
  <imagetype name="watermark">
  <filename>watermark/skype.png</filename>
  </imagetype>
  </state>
  

• Add to “/usr/share/mythtv/themes/defaultmenu/mainmenu.xml” an entry like the following

  <button>
     <type>SKYPE</type>
     <text>Skype</text>
     <description>Launch Skype</description>
     <action>EXEC /usr/bin/skype</action>
  </button>

Clearly you will have to change <your_theme> with the name of the theme you use (I use “Retro”), “watermark/skype.png” with the real location of your Skype logo and “/usr/bin/skype” with the location of your Skype executable (try with the command “which skype” if you do not know it), but everything else should work.

Restart MythTV and your new shiny Skype entry should appear at the bottom of the main menu. Clicking on it will stop MythTV and launch Skype. At first launch, maximize the Skype window with the mouse, then it will do it automatically. When you close Skype (for real, right-click on the systray icon and click close) it will go back to MythTV.

I am also fiddling with Lirc to allow to completely control Skype with the remote. It should not be too hard. I will update this post when/if I manage to do it.

The API is easy to use and to script but unfortunately there is often no list of symbols available to start with. Those lists exist on the sites of the markets (e.g., here is the one for NASDAQ) but they are hard to piece up together.

Luckily, with a could of lines of bash it is possible to create one from the list of components of the NASDAQ or S&P500. The following is the code necessary:

for p in $(seq 0 50 2739); do
    wget -O - "http://download.finance.yahoo.com/d/quotes.csv?s=@%5EIXIC&f=sl1d1t1c1ohgv&e=.csv&h=$p";
done >> /tmp/nasdaq.symbols


This script will loop in step of 50 from 0 to 2739 (numbers of element in the components list of NASDAQ) and downloading the list of symbols in each page and concatenating into the file /tmp/nasdaq.symbols.

The URL to download can be found at the bottom in pages list this one and you just need to substitute the &h=50 with &h=$p, which is the iterator variable.

They use GPS/time data sent back from people using Google Maps Apps on their phones (iPhones, Blackberry, Android, …).

At regular intervals the phone sends to Google its GPS coordinates and with simple calculations they can figure out how long it took to you to move from point A to point B.

If that happens to be on a road, and your speed seems to be the one of a car, they can use the data to estimate the traffic on the road, comparing the current data with historical ones.

Since wireless networks are pretty popular, and the combination MAC/SSID is unique, associating those with the car’s GPS coordinates allowed Google to create a pretty detailed map. This map could then be used to figure out your coordinates given the MAC/SSIDs around you. The technology is generally called Wi-Fi Positioning System (WPS).

A possible use of that is the Google Maps application. If the device does not have a GPS, it uses GSM cells triangulation (cells coordinates have probably be obtained in a similar fashion) to figure out the location. While it generally works, it cannot be very accurate and often has a 2 miles range approximation. However, if some wireless networks are detected in the surroundings, they can be used to produce a much better estimation of the location.

These data are probably also sold/used by the Automatic Geo-Tagging feature of the Eye-Fi memory card. Not surprisingly, those cards are sold in promotion with Google Picasa.

Simple word-based classification techniques like Naive Bayes will do the trick but require a very big training set especially for the foreign languages. Even if in the last years memory and processing power became less and less expensive, they are not free and especially in a startup you may need to optimize every single function.

For this reason, you may want to consider characters classification. In its simplest form, you just want to compute the frequency in which alphabet characters appear in each language and then compute the distance (e.g., geometric distance) between the text and your models. The memory requirement of this solution are very small (i.e., a float for each of the 26 letters) and CPU can be easily bounded as well (e.g., you can stop after N characters of the input text) trading off precision for speed.

If that is not enough, n-grams of characters (e.g., sequences of 2 or 3 adjacent characters) will probably work even better but require an higher memory footprint.

The following graph shows the frequency of the alphabet letters across the 5 most common European languages. For some letters the difference in usage is pretty high, e.g., the letter “A” is used twice as much in Spanish than in German while “H” is frequently used in German and English but almost never used in the other languages.

Unfortunately, after having installed it on Apple TV v3.0.2, it is not possible to play any TV Show due to an outdated Flash Player. No worries, here is how to unlock the Flash magic:

1. SSH into your Apple TV
   ssh frontrow@<ip_address>
2. Go into the temporary directory
   cd /tmp
3. Download the latest version of Flash Player for OS X
   wget 'http://fpdownload.macromedia.com/get/flashplayer/current/install_flash_player_osx_ub.dmg'
4. Mount the DMG file
   sudo hdiutil attach install_flash_player_osx_ub.dmg
5. Go into the Internet Plugin Directory
   cd "/Library/Internet\ Plug-Ins/"
6. Install the new Flash Player
   sudo pax -r -z -f "/Volumes/Install\ Flash\ Player\ 10\ UB/Adobe\ Flash\ Player.pkg/Contents/Archive.pax.gz"

Now restart your Apple TV and enjoy watching movies on Boxee (if the CPU is powerful enough).

In this guide I assumed you already unlocked your AppleTV using a Patchstick, which will allow you to install Boxee and other programs as well as loading an SSH server to allow remote connection. You can easily create a Patchstick using atvusb-creator.

Since the performances of my loyal Dell x300 are disk bounded, and I had success improving them using SquashFS last year (unfortunately it’s read-only), I decided to do my own review of Reiser4: I downloaded the source of my kernel (apt-get source linux-image-2.6.31-20-generic), its build dependencies (apt-get build-dep linux-image-2.6.31-20-generic), applied the right Reiser4 patch, copied over my current .config file (disable the DEBUG symbols from the kernel or your image will be huge) from the /boot directory and a couple of minutes later I was already running make-kpkg.

The goal of my first test was to compare Reiser4 with other well-known choices. I created a partition of about 5Gb and copied over my /usr directory (3549986832 bytes) using rsync (-aHAX) for each file-system.

Using Reiser4 the CPU utilization was definitively higher, but the disk throughput increased significantly (+53% versus ext4). If the ccreg40 plugin is used for file creation instead of the default reg40, Reiser4 offers many additional options on how the files are laid out, if and how are compressed, and which compression/encryption algorithm is used. I tested the impact of some of its compression schemes (conv, ultim and force) and the two compression algorithms (gzip1 and lzo1).

I was expecting the best compression performances from LZO1, but instead got the highest throughput (+70% respect to ext4) at the expense of an higher CPU utilization (+70% respect to Reiser3):

Using GZIP1 lead to the highest compression ratio (-48% vs. ext4) but a lower CPU utilization (-19% respect to LZO1) and a lower disk throughput (+42% over ext4):

The results can be difficult to interpret: LZO1 seems to give the maximum disk throughput but there is less CPU available for applications and the files on disk are bigger (thus there is more to read), on the other hand, GZIP1 offers lower disk performances but there is half of the data to read from the disk.

To help me decide, I devised a comprehensive score (lower is better) which takes in account CPU, disk utilization and throughput, assuming that they all have the same importance:

    score = (max_throughput / throughput) * (cpu / min_cpu) * (size / min_size)

Using this formula the winning configuration is GZIP1 with “ultim” (1.38). With default options enabled Reiser3 and Reiser4 obtain a score of 2.82 and 2.63 respectively. Slightly better GZIP1 with “force” (1.79) and “conv” (2.59). LZO1 is best using “force” (2.59) instead of “ultim” (2.80) or “conv” (3.25).

Using Reiser4 with GZIP1 and “ultim” should provide a 42% increase in disk throughput and reduce by 48% the amount of data to read, making my disk about 2.95 times faster. Let’s hope its true.

Considering one or two Tweets per day from each, plus 10 or more from accounts like CNN or ABC, it’s reasonable to think that you would have to look at 250 messages per day. And if you follow more people, or automated services, this number is even higher.

Who has the time to read all this? I doubt you will keep looking at your Twitter stream the entire day looking for those few gems. And if you do, most days you will learn that Bob is sipping a cappuccino and Jane bought new boots, instead of something really useful.

So I wonder: why Twitter clients (e.g., TweetDeck, Seesmic, …)  do not learn from  news products (e.g., Google News) and start adding classification, cluestring and raking of followers/tweets?

Classification would help probabilistically flagging tweets I may care about (e.g., technology, search, …) from the ones I do not (e.g., you are watching Lost, eating a pizza, …) pretty much like the spam filters do in modern email clients.

Clustering will put together all the tweets/discussions about the same topic (e.g., comments on the new movie of Bruce Willis) so I do not see them scattered in pieces here and there and I can quickly understand what is the general opinion on it.

Ranking could then take advantage of both classification and clustering, understand who I care the most among the people I follow, and rank the tweets in my stream accordingly.

With 41 Million tweets per day (39% containing link, for the majority spam) we could already take advantage of smarter Twitter Clients.

Nowadays, the Internet is a platform and technologies like Ajax and Flash reduced (perhaps even canceled) the gap between online and local applications. Most of the people who use Internet every day, I am sure, do not know the difference between email, Outlook, Gmail or Facebook Messages. In the recent years many providers are even pushing for Web OS, with applications (e.g., Excel) and storage somewhere in the cloud.

Tech people (2% at most?) know that it is a bad idea. The rest of the world (98%) will almost not even notice the difference. After all, their spreadsheet looks the same even in a browser.

Today, I read about Facebook’s idea of launching their own email platform: it is a genius idea.

Just a few days ago Facebook reached 400 Million users, and most of them log in every day. They definitively take a look at their Walls, the homepage, and check their messages. Facebook does not have to do anything more than setting up a few SMTP/IMAP servers, tell everybody that they now have an email <username>@fbmail.com and the deal is done.

I am sure that almost everyone who managed to create and use a Gmail account is on Facebook, so why bother checking both? Just sync  their address book the first time and goodbye Gmail. People do not even know what are the 8 Gb of space which Google is giving them for their emails, they do not use labels or stars, they not install addons nor use the IMAP capabilities..

For Facebook, this is a great move. They will be able to look into your email stream and figure out what your interests are to improve their targeted advertising. You will spend even more time on their site. They already managed to keep everybody logged in through the chat, and now, with the introduction of email and a better search experience (a tailored web search), people will have no reason to leave.

They are already the biggest photo sharing website of the world (Flickr who?). They are probably the biggest “forum” site of the world. Now they will also conquer email.

Good idea. Very good idea.

Things get especially messy with multi-level hashes, since you are obligated to pre-declare each level as shared.

Luckily, there is a way to make things easier. If you upgrade threads::shared to version 1.32 using CPAN and can afford to waste some memory for a little, you can create your objects normally and then create shared copies of them using shared_clone().

This function will recursively traverse the object, create a shared clone of each element in it, and return you a nice reference which you can pass around to your threads.

At that point, to save memory, you can undef() the original object and keep only the clone.

This works great and flawlessly for read-only objects but it will still require some caution when you want to modify or add/append data to them since they need to be pre-declared as shared.