On Aug 19, 2004, at 4:25 PM, Dave Watson wrote:
there are a lot of ways you could cause the server to crash that
cannot be cheaply checked (i.e. doing a graph analysis at each
operation), so none are checked for. Perhaps this is reckless but
that is the way it is.
Preventing the removal of node zero when an n_free is received causes
a decent performance hit?
That is not what I said.
rephrasing:
Because there are a lot of graph operations that could cause the server
to crash that cannot be cheaply checked, no graph operations are
checked. Not even simple ones. Perhaps this is reckless but that is the
way it is.
