I am reading the current, April 2009, issue of Mechanical Engineering, and there is a nice article on mass customization, "Democratization of Manufacturing". The article includes a description of emachineshop.com, a firm that accepts uploaded cad drawings, and then automates the production of parts from the drawings for the customer.

Does anyone have any experience with this? This is another example of where I believe technology will enable the replication of complex parts for our restorations. A reprieve for Winton engines?

I don’t think this represents any real breakthrough in manufacturing. It appears to be just a unique business package where they supply the design software and a relatively streamlined process. But in the end, the process of making complex parts is fraught with problems. I would not further tempt fate by using this streamlined, design-it-yourself business model. I think that expecting the customer to use their free software to model complex parts and get the proper result is a bit much. Any complex engine parts could be made by hundreds of competent shops whether or not you provide a cad solid model. All you need is detail drawings, tolerances, and the metallurgy.

Well there's CNC, and then there's these guys.

To me, CNC is indispensable. It has no limits on complexity, is reliably repeatable, and you can prototype with trial and error (make sawdust til you get it right, then make chips.)

These guys... I expect they excel at repeatability. Maybe you can even do a LuLu / Cafe Press model, where you design a part and tell your customers "Go order it from them" and get kicked back a royalty. That would be COOL.

Complexity is obviously unbounded, but... the more complex, the more you depend on the "trial and error" process - and that's where they stink. For each trial cycle, you must pay full boat and wait for the UPS-man. Worst of all, they use their wacky custom software -- which means you can't prototype at home on your inexpensive wood CNC and send them your final. I suspect the software facilitates a) avoiding many pitfalls and traps in CNC, and b) generating price quotes.

Edit: Tried it!
Just for reference, last year I joined TechShop and made a sheetmetal part for a project. 13 holes, 6 square; two bends. Quantity 60. The CNC plasma cutter made this easy, but there was a lot of trial and error getting the holes in the right places. Cost: $125 techshop membership, $50 machine training, $400 in steel.

Anyway, I fired up eMachineShop's software and modeled that same part. Took only about 15 minutes despite learning the software totally cold. I pushed "show me 3-D model", and it was perfectly right first time. Then the fun began: costing. The software can quote prices and delivery dates. It's doing this offline without internet connection, so obviously it's wild-guessing delivery date.

What's more, I was able to play "what-if" with the costing. Materials, processes, quantities... change something and hit "cost", see what it does to the price. Very impressive. Some of the price differences were dramatic. The software also had a few suggestions for reducing cost. Shipping was included in the price quote. It looked like somebody put a lot of love into this program.

I couldn't find a way to print a proper drawing showing all dimensions (as for instance SolidWorks can)... which would make it tough to compare what you drew in their tool to the official drawing.

What I find more intriguing is future possibilities this reflects. For instance, a standardized way of modeling a part, so that you're giving your machinist a computer file instead of a paper drawing, that you could take or send to any machine shop anywhere and they could cost it or build it. And how about standardized costing - that's something the auto repair industry already has. And most interesting of all is the possibility of some parts being essentially "make on demand", with a "Cafe Press" style royalty model.