I find myself running into the following situation writing C# programs
that have controls and components that have custom designers associated
with them:

1) I need to write code that uses non-redistributable assemblies (e.g.,
EnvDTE) at design time, but not at run time.

2) Referencing Types from the non-redistributable assemblies in the
same assembly as the redistributable Types causes a runtime exception
on my client's machines (because they do not have the
non-redistributable assemblies installed).

I "solve" this problem by segregating the code sections that use
non-redistributable Types into a separate assembly. As long as the code
that references the new "non-redistributable wrapper assembly" (sorry,
it's late, and I can't think of a clever name) never gets executed, the
previous run time exception is avoided.

While this works, it feels kludgy (for one thing, I have to
redistribute the "non-redistributable wrapper assembly", even though if
it were ever called it would cause an exception).

What's a better design pattern to use?

- Mark