There are two ways you can use MobiLink to sync to a database you don't have DDL access to.

The most straight forward is to use the -cs server option. This lets you put the server's system tables in a different database than the consolidated. In this case you would have an SA database you control which contains the ML system tables and the MobiLink server would have a -cs switch specifying an ODBC connection string to connect to that SA database, and a -c switch specifying an ODBC connection string to connect to the Oracle database. The big caveat with -cs switch is that it requires server to use the Microsoft Distributed Transaction Coordinator, which can be slow. The -cs switch was added in version 11 and you can use a version 11 server to sync with your version 9 SA database. See http://dcx.sybase.com/index.html#1100en/mlserver_en11/ml-syncserver-s-5359674.html*d4e2495

The alternative is a bit more complicated. You can use the MobiLink direct row API to inspect the upload stream and do whatever you want with it. As with the first option, you'll need an SA consolidated for ML to use. See http://dcx.sybase.com/index.html#1100en/mlserver_en11/handle-uploaddata-ml-scriptsref.html*d4e15667

There's an example from that page that will do almost what you want:

public void handleUpload( UploadData ut ) 
  throws SQLException, IOException {
  // Get an UploadedTableData instance representing the
  // email_addresses table.
  UploadedTableData addressesTable = ut.getUploadedTableByName("email_addresses");
  // Get inserts uploaded by the MobiLink client.
  java.sql.ResultSet results = addressesTable.getInserts();
  while( results.next() ) { 
    // You can reference column names here because SendColumnNames is on
    // Get the primary key.
    int pk = results.getInt("pk");

// Get the uploaded num_ordered value.
    String email = results.getString("email_address");

// The current insert row is now ready to be uploaded to wherever
    // you want it to go (a file, a web service, and so on).
    magic( pk, email );
  }

results.close();

}

This opened a JDBC result set over the data that was uploaded from the remote and then called the magic function for each row in it. You just have to write the magic function to do what you need done. For example, you could connect to the Oracle database and insert the row, or you could toss the rows in a file and later email them. If you also want to send a download to the remote, there is API for modifying the download stream. See http://dcx.sybase.com/index.html#1100en/mlserver_en11/ml-objectbased.html for the full direct row API docs.