On Mon, Feb 12, 2007, Tony Finch wrote:
On Sun, 11 Feb 2007, Claus Assmann wrote:
Checkpointing is "out" as it has (security) problems between untrusted
I don't think it has significant problems if you only use it for duplicate
elimination, i.e. without data checkpointing. Do you have any omissions to
As long as checkpointing is not a requirement for "per-recipient
data reply" but just mentioned as one possible solution to the
problem of duplicated mails, I'm fine with it. Requiring checkpointing
would keep people from implementing "per-recipient data reply" due
to the complexity of a possible checkpointing implementation.
The RFC should simply state that the problem mentioned in RFC 1047
requires that senders and clients at least properly obey the timeouts.