Well, this is what I could do in that (or almost any) case: 1) Setting features, key points, points to avoid, and goals of the project. 2) Choosing the toolch

Well, this is what I could do in that (or almost any) case: 1) Setting features, key points, points to avoid, and goals of the project. 2) Choosing the toolchain to use, for resolving the points at 1) 3) Coding the features with that toolchain Step 2, may involve: to read a lot (doc, rfc's, specifications, using search engines with keywords, search for "problems", etc) to look at other people's source code of similar programs, to try other people's tools and libraries and evaluate and compare them, to make integration tests, and to make Proofs Of Concept prototypes. The toolchain (programing language, user GUI, etc) comes with an ecosystem (I hope you don't aim at rewriting also your $lang libraries for validate XML or interact with devices), so you have to choose looking at both. To separate steps 2 and 3, use iterations, while you have resources. If you find an opensource product that full-fills all your definitions of point 1), except a few of them, you can evaluate to start by a patch and a pull-request to them, or a fork :P