Archive for December, 2009
Does British Airways train its New Jersey Attendants?
This year for my Christmas trip I will have the pleasure to fly oversea with British Airways. Although I have taken many of their planes in the past, it was always for short trips, and I never tested first-hand if their in-flight experience really justifies the usually higher prices.
Their reservation website definitively got better in the last years. It is faster and shows you the very convenient 3-days-around-your-dates tabled view first introduced by Orbitz. However, I do not appreciate that they make you pay $15 or $30 to pick your seat more than 24 hours before departure.
Today I received their reminder-email which suggested me to fill out an immigration form online (I am sure I will have to do it again at the airport but I am hopeful) and pick my seat (for the $15~$30 mentioned above). The most hilarious part was the flight details. Apparently, as you can see from the image attached, I will have “Responsive, highly trained cabin crew” on all my flights besides the one which leaves from New Jersey.
I guess the folks at MTV had their reasons when they recently decided to launch the reality series Jersey Shore.
Sockets Connection without Listen()
Today, invited by an apparently impossible condition that showed up on OneRiot‘s servers I pushed my good friend Alessandro to prove me wrong. And as usual, in about 15 minutes, he did.
The odd condition was the following: server A was listening on a port and server B was connected to it. We turned off server A and after a few minutes server B was connected to itself on that port, although it was not listening on it.
Further investigation on the RFC 793 (page 23), and 20 lines of C code written by Alessandro, showed that is intentionally possible to connect two sockets together without calling listen() on any of the two. It is called “Simultaneous Connection Synchronization”.
On our server pure chance is what generated the condition. Server B kept trying to connect to server A cycling through the ephemeral ports allowed by the OS until accidentally the source port was the same as the destination port.
From now on our ephemeral ports range will be outside of our server ports range. I strongly recommend you to do the same especially if you run multiple services on the same box.
PubSubHubbub: a 1987 idea with HTTP/XML and Peer-to-Peer Sprinkled on It
If you read tech blogs like ReadWriteWeb or TechCrunch you probably have heard of PubSubHubbub, a distributed publishing method recently announced by the Google’s folks: PubSubHubbub.
Tech bloggers are going crazy about it and wrote thousands of posts without really knowing what it is and who will benefit from it. It is one of the buzzwords of the moment and nobody wants to miss on it.
But anybody who studied Computer Science in college will probably remember the Publish-Subscriber model from some of the introductory classes. When simple poll models, in which who is interested in new data constantly asks for it, are too expensive or not scalable, everybody switches to a push model, in which who is interested in the data (subscriber) let the creator (publisher) know and will receive updates whenever there is something new. This was invented in 1987.
Add some HTTP/XML, sprinkle some ideas from peer-to-peer systems, and 30 years later you have PubSubHubbub.
Seriously, that is the idea. The publisher picks some hubs and let them know that it will be publishing something. When it has some updates, pings them (with an HTTP POST) to let them know. Once alerted, Hubs go fetch the full content and distribute it to all the subscribers who previously registered with the hub for that particular feed.
Yes, it will make your little blog scale better since people will crush the Hub and not your site (but isn’t it on WordPress/Blogger servers anyway, so why you care?), but how many blogs/publisher out there have this kind of problems? And if they are that important to receive so much traffic, shouldn’t they actually think about that as a business?
Finally, why isn’t anybody talking about the Hubs? The system is “simple” for publisher and subscribers, but who designs, runs and maintains the Hubs? If an Hub goes down all the subscribers lose the updates (yes, they can go to another Hub) so those systems need to be redundant and scalable (they have to download the content and distribute it).
The only interest one can have to create and maintain an public/free Hub (as Google has done) is to get an hold on the data. Instead of crawling millions of blogs (publishers) to check if they have been updated, they will let you know. At the same time, you will know who (subscribers) is interested in what, and as Google has shown in the past years, that is pretty handy information.
New Google Interface: Bubbly like Ask3D, but I like it
It seems that after the usual gazillion of user testing the folks at Google finally settled on a new shiny search interface. In short, I like it!
We have to admit, the choice of colors, the shape of the buttons, and the use of big text resembles a little the Ask 3D interface that we launched in Ask.com in June 2007, but I think the visual impact is very nice and clean.
The menu on the left allows you to narrow down your search (and I am sure it is sorted by what people use the most for each query) in categories like News, Videos or Images, explore more about the topic with some related searches or specify a date range for the results.
Although the interface changed they apparently abandoned the idea of “abusing” the #fragment in the URL to run the searches through AJAX (as some sites like Collecta are now doing to increase their SEO/page-rank) like they were testing a few months ago. I am glad of this choice, it was just confusing.
My only question is why are they leaving the top menu in the result page. It is clear that by then the user had no intention of logging in or clicking on the Gmail link, and if they do want to switch to a different search category it can be done more comfortably using the left menu. I would remove it.
Also the homepage has been updated with a more colorful and bubbly style, that matches the results page.
However, some users w
ill see yet a different (and much uglier, in my opinion) version of it, which resembles the old-style Google homepage but with no upper menus nor bottom links.Those links will fade in if you leave your cursor for about a third of a second in their areas. Apparently a lot of studies went into that, up to the length of the pause (in milliseconds!) after which the text fades in.
