Sanskrit, the secret of India's outsourcing success. 16

(But first...) Cross-post to sci.lang added, in the vain hope that someone there will be able to sort out this mess about Sanskrit being an ideal language for programmers to learn.

If you are writing inadequate and incomplete specifications, I would, with all due respect, suggest that you need to learn how to write better specifications, not learn another language. If you are not putting enough information in the English version, using Sanskrit will not suddenly add more information to your work.

Certainly not. The XP tests-first approach is an approach to building the software, not specifying it, and not even "task definition". In fact, writing the test cases is (depending on how you slice things) either a task in its own right, or part of the larger task of design and implementation.

To proceed "a priori", you must start, in effect, from nowhere, since you are reasoning based solely on theoretical considerations, and here you are starting from a position with plenty of documented evidence to support it. "A posteriori" would be a better choice.

You are likely to find that it will be very UNforgiving. My own experience of building super-late binding systems shows that when you rely totally on configuration to make the system work correctly, you become vulnerable to even the tiniest of errors in the configuration.

The entire system is coupled very tightly to the interfaces, and is vulnerable to the slightest change.

Provided that the extensions can fit into the base-level interfaces.

If you can avoid it, don't develop during requirement discovery. At the very least, do enough requirement discovery to be able to specify high-level behaviours, as well as enough detail in at least some part of the system that you can build something real. If the customer (whatever that means for your project) does not know enough about what he wants to be able to tell you, you can't build it.

-- SteveR

Humans are way too stupid to be dumb animals.