[PATCH 09 of 10] mq: mq.secret prevent mq changeset outside secret phase

[Matt Mackall]

On Tue, 2012-01-31 at 19:15 +0100, Pierre-Yves David wrote:
> > > B) if B becomes public: we have a public changeset handle by mq that can
> > >    rewrite it. this break expection of the public phase.
> > 
> > Reconcile these two statements for me:
> > 
> > "mq will fails to qpop or qref B"
> > "if B becomes public [...] mq that can rewrite it"
> 
> If D don't exist but B is on the server, B is public and the safety above don't
> applies. This probably qualifies as a bug and mq should refuse to qpop or qref
> public changeset too.

Absolutely.

> > Remind me why we're doing this whole phases thing again. I thought it
> > was so that when a changeset became public, we would recognize that and
> > cease trying to modify it.
> 
> > This patch seems entirely at odds with that.  It's basically covering your
> > ears and saying "no no no, no one knows my secret".
> 
> This patch is more about: You asked me to keep this secret but other people are
> already aware of it. I'm a stupid machine, please tell me what to do!

But there is no semantic ambiguity in this case: a secret that everyone
knows is simply not a secret. You might not like that your secret has
been made public, but that has no relevance.

-- 
Mathematics is the supreme nostalgia of our time.