On Tue, Feb 11, 2003 at 03:01:31PM -0500, Bruce Lilly wrote:
WJCarpenter wrote:
km> it takes a very long time before support is no longer required -
km> people still want to be able to read old email or news messages.
I'm a little confused by this line of reasoning. We're talking about
wire formats, not storage formats, aren't we? What has reading old
anything got to do with protocols?
To name just one case; forwarding of old messages, either via the
infrequently used 822/2822 Resent- header fields, or by MIME
message/rfc-822 encapsulation, or by cut-and-paste.
Ok, so assume for a moment that while we can (someday) prohibit the
use of the older format on usenet, reading tools probably have to
be told that they must decode the older 7 bit format forever, in case
they need to read an old article. Forwarding is not an issue, as
tools can be told to map the old 7 bit encoding to the new one in that
case. Even though that has risks they are fairly minor, and this will
be an increasingly uncommon event.
Note as well, that actions like these are why I proposed the new
semantics for "Distribution." Once those semantics are in place, you
can be assured that no article with a specific Distribution will be
processed (at least by conformant software) at a site or gateway that
does not subscribe to that distribution.
It thus becomes possible, very quickly, to define subsets of the nets,
which, early on, embrace the 8 bit format. This requires that people
understand in writing mail gateways that they must not forward to mail
any article that comes with a distribution not explicitly subscribed to,
in this case it may not even be safe to do so.
Because distributions are expected to leak in relaying, just not in
local data storage and reading, all relay servers must be ready to handle
the new format before this can be done, however.