diff --git a/src/tricera/concurrency/CCReader.scala b/src/tricera/concurrency/CCReader.scala
index beef2b0..948e83f 100644
--- a/src/tricera/concurrency/CCReader.scala
+++ b/src/tricera/concurrency/CCReader.scala
@@ -163,6 +163,32 @@ object CCReader {
 }
 
 ////////////////////////////////////////////////////////////////////////////////
+trait ProgramBackTranslator
+object DummyBackTranslator extends ProgramBackTranslator
+
+trait CCASTPreprocessor {
+  def translate(prog : Program) : (Program, ProgramBackTranslator)
+}
+
+// 0) Compile visitors in ccparser, currently only the ones in Absyn are compiled.
+//    Look up Princess SMT-LIB parser, which had the same problem which is now fixed.
+//    -- Do the same for the ACSL parser as well.
+// 1) Do type-inference in first pass (could be separate visitors or the same)
+//    Extend EVar class and create a new AST after the pass.
+//    We could also extend the grammar, e.g.,
+//      EvarWithType. Exp17 ::= "$$" CIdent ":" [Declaration_specifier] "##";
+//      This has the advantage of making all generated visitors know about this.
+//    Challenge:
+//       -- ComposVisitor will revert back to the old AST -- won't know about
+//          the augmented classes.
+//       -- Solution: Extend ComposVisitor so that it knows about the new types.
+// 2) Detect pointer types, rewrite call sites.
+//    -- At each call, i) if it is a stack pointer rewrite using a new function
+//       name and global variables.
+// 3) Create copies of functions for stack & heap pointers.
+// 4) Add back-translators.
+// 5) Stretch goal: output contracts in ACSL AST, and implement preprocessors &
+//    back-translators using this.
 
 class CCReader private (prog              : Program,
                         entryFunction     : String,