Franz' ACL 5 has an interesting 'gotcha' related to backquote syntax.  Consider this form:

`(:a ,x :b ,y :c nil)

The list returned is not entirely fresh.  In fact, the last two conses comprise a constant list.  Therefore, if you destructively alter the last element, all lists generated from this form will be updated.  Ouch.  I am not sure whether this behaviour conforms to ANSI or not (I haven't looked).

I discovered a problem with our ORACLE.C module while working on ITEMSEQ.  If you use bind variables in an SQL statement (e.g. an INSERT statement), our call to OEXEC hangs indefinitely.  At first, I was concerned that the problem existed only in LISP.  But I wrote a small C program that called ORACLE.C directly, and it exhibited exactly the same behaviour.  It made no difference whether blocking was turned on or off.  When turned off, OEXEC always returns the 'would block' error code.  I am stumped as to why this behaviour occurs.  Perhaps it has something to do with the fact that we are using the old, old, old, OCI interface.  Oracle may no longer support that interface.  Bind variables do seem to work for queries.

When connecting to DB2 using the JDBC driver, an additional property gets added to the supplied java.util.Properties object.  The property is 'LANGUAGE=C'.  I don't think the DB2 JDBC driver is supposed to alter this parameter.

Hey, some good news about ACL5!  (About time, too.)  ACL5 allows CLOS objects to be patched!  I tried loading a FASL into an image that contained both DEFCLASS and DEFMETHOD, and it worked.