\documentclass{article}
\input{6824-preamble}

\begin{document}
\psetnum{7}
\date{2005/03/01}

\begin{pset}
  Porcupine ensures that the failed node receives a copy of $M_2$
  because the coordinating replication manager (which is the node that
  received the message) transmits a copy of the update to the failed
  node. The node does not receive a copy of $M_1$ because two update
  messages have been transmitted, one creating the message and one
  deleting the message; when the updates are transmitted from the
  replication coordinator to the failed node, it will not commit them
  to disk before processing both updates, ensuring that $M_1$ is not
  created.
\end{pset}
\end{document}