Error Handling And Propagation Of Errors To Caller

With the new features of SQL Server 2005 for error handling (TRY...CATCH blocks), how are you propagating errors back to the caller? For example, lets say we have 3 stored procedures:
dbo.usp_UpdateSomeTable1
dbo.usp_UpdateSomeTable2
dbo.usp_UpdateSomeTable3

Let say some application calls dbo.usp_UpdateSomeTable1, in turn dbo.usp_UpdateSomeTable1 calls dbo.usp_UpdateSomeTable2 and in turn dbo.usp_UpdateSomeTable2 calls dbo.usp_UpdateSomeTable3.

Now if dbo.usp_UpdateSomeTable3 generates an error, how do you handle propagating this back to the caller?

I envision encapsulating the contents of each procedure in a TRY...CATCH block like so:

BEGIN TRY
...do some stuff
END TRY
BEGIN CATCH
...handle errors - whether generated from our own RAISERROR statements or by the database engine.
END CATCH

Now my problem is I would like to capture all the error variables and toss them back to the caller and keep sending that information up the stack. So far my attempts have been pretty unreadable and end up being just a cluster of text.