I was just wondering if it's easier to start building upon the tiles application than starting to create a new one from scratch, ie. webtiles.
Well, no, it isn't.
The client is the easiest part here. The hard part is choosing which data to send, when to send it, in what format, and then making crawl do that -- basically what you called "just synchronizing the data". For example, the server can't just send the whole map after every step: that would take far too much bandwidth. So it needs to keep a copy of the map as it is displayed on the client, and then compare what changed and just send that. Similar things need to happen for most other data.
Also, the webtiles server already reuses parts of the tiles interface (tile picking, for example), and large parts of the client are modeled after the tiles code.