I view existing engines as an excellent starting point.

Technology progresses when we build off of the work of others. Constantly reinventing wheels can teach a lot, but in the terms of actually accomplishing greater things, what does it really do?

In a number of cases these engines are the products of years or even decades of work...

In the end the question must be what do you hope to accomplish? If it is to learn how to do collision detection and response then by all means implement it yourself, but if you wish to make games then using existing technology is a fantastic idea.

If nothing else in several years or with time why not contribute to an open source existing physics engine? That way you could advance the technology and not start from scratch.

Just my two cents.

I view existing engines as an excellent starting point.

Technology progresses when we build off of the work of others. Constantly reinventing wheels can teach a lot, but in the terms of actually accomplishing greater things, what does it really do?

In a number of cases these engines are the products of years or even decades of work...

In the end the question must be what do you hope to accomplish? If it is to learn how to do collision detection and response then by all means implement it yourself, but if you wish to make games then using existing technology is a fantastic idea.

If nothing else in several years or with time why not contribute to an open source existing physics engine? That way you could advance the technology and not start from scratch.

Just my two cents.

That is true too.