Comment 6 for bug 605017

I found a case which is better served by patch in comment #1: if some section alters an extended section during initialisation, my fix would hide that alteration, while patch #1 would make it available (because extentions are always walked through entirely, while my patch caches the result of first walkthrough).

I cannot think of a valid case for such alteration, but nevertheless patch #1 pushes less assumptions to the code than mine, hence it is better.