Monthly Archives: December 2007

Is the fairydust falling from the Googleplex?

Dare is right. I have also thought that the Googleplex seems to have been shedding some of its geek lustre over the past few months. Don’t get me wrong – I think some of the fabled infrastructure they have (GFS, MapReduce, Big Table, etc) is beyond world class. Even if it did not exist and was purely vapourware, it has done the world a massive service by opening peoples eyes to what can be done with massively distributed systems. I cannot imagine living without my Gmail account. It wouldnt occur to me to use any other search engine.

But there is a nagging doubt. Is it really the hallowed turf of engineers or have the MBA types started to ooze in? Take the Knol. I don’t get it. There are a billion and one different ways for people to put content on the web. I am not saying that our current tools (blogs, wikis, myspace) are the ultimate evolution of content authoring. Clearly they aren’t. But what does the Knol actually add? As Udi says: “it is still just a web page“. The only thing I can see is that it is yet another space for adsense to pop up. Adsense was a brilliant revelation and clearly works. But you get the feeling that Google, for all its glitter and fairy dust, is actually a pony with only a small repertoire of tricks (albeit very, very powerful).

I have no doubt that they will come up with some more truly, truly amazing ideas. Having a brain bank the size of a small galaxy makes that an almost guaranteed cert – you would back it at almost any price. But it doesn’t surprise me that people, who once would have sold their mothers hind legs to work there, are now looking elsewhere.

PS: I am not sure about Dare’s other horse. I can see why the money types are so enthralled with Facebook but it is a big WTF? to me. Anyone that can put out Beacon scares me.

PSS: Guess that is two companies I won’t be getting offer letters from then!

Startup heaven or megaprise tar pit?

Another day, another step towards Amazon’s play for global domination of tieing in startups to their infrastructure. The latest kid out of the blocks is Amazon Dev Pay – a one stop shop for handling customer account and payment functionality. Anyone who has had to deal with this kind of stuff knows that handling it well is both critically important to a business and also, generally, a sideline to its main activities. It is also quite hard to do right. The question is: should you use it?

I think it is going to be very interesting to see what happens to those startups that have successfully bootstrapped themselves up on the basis of Amazons infrastructure. They are going to be in the rather precarious position of having (hopefully) a very successful business on their hands and also being almost entirely dependent on someone elses infrastructure. I know Amazon made a big step forward a few months ago with the announcement of SLAs but you have to wonder how tightly coupled a lot of these companies actually are to AWS.

As it currently stands I would imagine that many of these companies have no desire to leave the safety of AWS. I also have no reason to believe that Amazon have any intention of causing them any harm. Indeed everything I have seen and heard from Amazon indicates to me that they know that what is good for their customers is good for them and they would actively try to do the right thing. But you have to wonder what will happen when one of these startups becomes a megaprise in its own right and starts to cause tension in the relationship. Migrating out of AWS is going to be an incredibly difficult proposition for many of these companies. What currently seems like startup heaven could well become a megaprise tar pit.

Tim Bray was right recently when he said that for a startup “Getting started should be free. Also, it shouldn’t take more than a few days.“. Amazon clearly subscribe to that mentality. I wonder whether they also subscribe to one of Tim’s other points: “Try to lock them in and they’ll walk away. Put it another way: make it easy to walk away, and they’ll come back.”?

What would the Amazon legal team do if someone created an opensource implementation of S3 or SimpleDB in a similar manner to Hadoop? Would they allow an open source implementation that even cloned the APIs? Whilst the technical services have the luxury of making use of Amazons vast compute scale and would be, basically, impotent for the vast majority of users what about the financial services such as Dev Pay? Clearly this is a service in the same right as all the others, and it is not clear that it is not reproducible on a much smaller scale. Will Amazon try to lock you in or make it easy to walk away?

Amazon DB and Erlang

The world and it’s dog is jumping on the Amazon SimpleDB announcement. Interesting to note that apparently (big caveat there!) it is built on Erlang.

I am keen to see how the DB heads respond to this. I also want to see how how long it takes for people to get their heads around the eventual consistency model it employs. The use of an eventual consistency model is often a critical factor in being able to scale any ultra large system but the concept scares many people to death (and thats only the people who realise what it means!)