These days, I can't go more than 10 lines of code without having to call some API* that requires me to be online.
* Examples from just yesterday: Amazon, Twitter, Facebook, PayPal, GData Contacts, GData Maps, Bing Maps, YouTube.
2. Code for a day offline with the docs.
3. Test your code (online).
4. Submit a ton of bug reports for the docs.
5. Yay, improved documentation!
7. Get outstripped by competitors.
When you say you "download/clone", are you talking about just the docs, or do you actually clone the API locally?
For example, create a local, mock Twitter OAuth API so that you can program against the Twitter API without being online.
I'm genuinely curious. I like the idea of going offline, but it seems like it's not worth the effort if you're working on applications that are codependent on network APIs.
I wrote this to help: https://github.com/bradjasper/focus
For instance my work last week relies on an API that requires a VPN connection AND a windows machine.
I tested and mocked the API on windows, and now work offline, without the VPN nor Windows.
For instance ... if you need to read the Amazon AWS's docs, maybe you're better off searching for and using something like Boto. If the Boto API is too complex to memorize, maybe you're better off constructing a Facade around it that you can remember without going for the documentation every time you want to open an S3 bucket.
Just saying ... lack of easily accessible documentation leads people to do strange things :-)
If you're building a Facebook app, for example, you need to have Facebook available at all times or it just won't work.
I just did a PayPal API, and the docs were pretty bad/inaccurate, so we had to make frequent test calls to see what the real world behavior actually was.
It gets worse. There are several situations where the docs say one thing, the sandbox behaves a different way, and the live API behaves in a third, different way. It's a mess.