Wow, that sounds like a fun project and assured income for the rest of your life! Crafting initial tools from flintstone, putting together a computer without the aid of industrially premade parts, I'd reckon it'd take you several decades at least before you even get to the software stage. By the time the project is finished the singularity will have come and gone and rest of us will have to bring up the Wikipedia page to remember what a shopping-cart even is, but you'll have been in bread and butter for all that time.

Snide jokes aside, unless the client has some really concrete technical reasons for not building on existing software, you should dissuade him. A good tack (in my experience) is an emphasis on maintainability. If you get hit by a bus tomorrow (Dog forbid!), the client will have a much easier time finding someone else to support a system that is written using well-known components than a completely hand-rolled one. No matter how clear-structured your own code is, a large user- and developer-base with forums, wiki-docs etc. will always be more help in finding and solving problems.

If the client does have concrete reasons, find out what they are and do some research. Maybe a solution exists that your client just hasn't come across? Or you may have to write a bit of "glue" to help fit together his notions with an existing solution? This is all a bit vague, but you haven't really given any concrete problems for which one could suggest a solution.