This is a good question, and one worthy of an *official* response.
It would be naive to provide a guarantee that mbed.org will never disappear. We are trying something new, things happen outside our control, and we can't predict everything about the future. There is always a risk.
So, if it did happen, our intention would certainly be to provide everyone with a way to download their code, and an offline alternative to the online compiler.
With that doom-and-gloom out the way, I'd like to highlight that we are very committed to this project - and by this I mean both the people working on it and the increasing number of companies helping to support it.
You may not know this, but we actually started the project in it's infancy late 2005, and were already running it as a service by 2006 for some very early alpha testers. Yet we only released it generally this October 2009; why did it take so long?
Firstly there is the obvious technical reasons - we had it working as a prototype in 2006, but that is only 10% of the way to a product. So there has been a lot of development going on, but this is only one aspect.
Secondly, we wanted to test it worked as a concept. So we grew our alpha test base initially with friends/colleagues, then a beta test base by openning it up to people who somehow found out or were introduced to it. All this helped us improve and learn how to run mbed, whilst properly testing if it actually worked and was a good idea.
Finally, and perhaps most importantly, we wanted to ensure it was commercially sustainable. We've seen many great things come up over the years that ultimately fail as they couldn't sustain themselves - either they never got popular and so died, or became popular but then couldn't afford to support themselves, and hence had to be shot! We set this up as a robust model that if successful, enables ongoing investment in itself.
For example, I've seen some moans by people who don't like mbed that $99 (or $60 atm!) is a lot for a board. Well, I'd say it is a bargain! Even if you only consider all the component parts on the board, manufacture, packaging, logistics, distributor margin etc, its a pretty good deal. But there is also a bit in there to go towards "mbed the service" - that is what the license key represents. For your $99, you're also getting access to the mbed compiler and the mbed world. Put it another way - in return for some of your hard earned cash, we provide tools and a supporting website for your programming pleasure.
We also managed to get it backed by some great companies. ARM is obviously one (that is where we work), NXP is another (the microcontroller the first mbed is based around), distributors like Digikey, Farnell, Future, Mouser (see http://mbed.org/order) who are doing more than just selling it. And we'll be announcing some more in time. All this support and alignment really helps make it sustainable.
So by the start of 2009, we were satisfied we had all these aspects covered. Yes it took a while and much blood, sweat, tears, but we weren't going to commit unless we were really confident.
In conclusion, there is always a risk with something new. It basically comes down to a risk-reward decision that we can't answer on your behalf. If you can't afford the risk, go with something else. If you think the upside outweighs the potential risk, come join in! Hopefully this explanation highlights we've done a lot to mitigate this risk, and will help you make an informed decision.
Just got my mbed and I've started futzing around with it. Great stuff!
But what happens if mbed.org disappears? Is the online compiler and its attached libraries available anywhere for download?