The opposite of over-engineering is under-engineering. Creating the least viable product. The most minimal, simple implementation that can do what you want it to do. Counterparty is this solution for the distributed exchange/Bitcoin 2.0 idea. It works, today. It does so by leveraging the existing Bitcoin network and protocol, which allows the developers to devote their time to getting the important parts right. How much would the Counterparty codebase increase in size if it required its own blockchain? I'm betting it would be somewhere in the triple to quadruple range. Counterparty will continue to work for at least the near future. If a day comes when it is made not to work by outside forces, Counterparty will change in order to work.
Actually that's incorrect: Counterparty and Mastercoin both have a significantly larger codebase by
not requiring their own blockchain. Of course when I say "larger codebase" I mean the code required above and beyond the Satoshi codebase itself. Freimarkets is an example of a system that took the opposite approach of reusing the Satoshi codebase and requiring its own blockchain, and by doing so it got to reuse blockchain-related code from the Satoshi codebase that Counterparty and Mastercoin had to implement themselves. Of course, what they implemented is more simple than reimplementing blockchain code from scratch, but their approach is definitely not the "least-work" way of doing so; Freimarkets is.
But if "what you want it to do" is to be secure against attackers, then the minimum viable product has to be an embedded system as opposed to independent or merge-mined and you're stuck writing embedded-consensus blockchain code.