I have reviewed and written some ethereum contracts for test purposes and i have two questions.
After you write your first contract, you will see that (in most cases) it's basically useless without an interface. You need an interface to interact with the contract. And right now if you want the large public to be able to interact with your code, there is only one option. Write a web interface, upload it and invite users to use it (
http://www.augur.net/).
Basically you have a centralized website that allows users to test your decentralized application. While the app is unstoppable, the web site is as centralized as it can be.
Question number 1 : What is the difference from a centralized php / mysql app running on domain www.xxx.com and a decentralized app that can be used only by accessing domain www.xxx.com ? Because i don't get it.-------------------
Second. After you upload a contract, the contract will stay on blockchain forever. If i upload 10 big contracts, and i quit or loose my private keys, those contracts will be there forever. For scalability it's a disaster.
Question number 2 : What stops me from writing 1000 huge contracts every day, abandon them and inflate the blockchain ? Those contracts will never be executed, so there is no costs for me excepting the initial "upload" cost. This could be a real attack vector right now, because tons of contracts were used last time 30-60 days ago 
