ietf-nntp MODE READER

[Russ Allbery]

Clive D W Feather <clive at demon.net> writes:

>   Servers SHOULD NOT alter their behaviour depending on whether or not
>   the client has used the MODE READER command. Any authentication SHOULD
>   be done using the identity of the client (e.g. its IP address) or
>   separate authentication extensions. Server authors are encouraged to
>   remove any side-effects of MODE READER.

"SHOULD NOT alter their behavior" is not existing practice.  Again, it may
be useful here to get a clarification of the working group's scope.  I'm
worried that if we try to start solving all the problems in the protocol,
this is going to take forever.

Anyway, common practice in my experience is moving quickly towards running
a reader server and a transit server on separate ports.  If we *have* to
try to deprecate MODE READER, can we please use language that suggests
that approach rather than using language that makes it sound like you have
to provide all the functionality of a reading news server in your transit
news server if you split them apart as an optimization?  I realize that
the above text doesn't specifically say that, but I think a lot of people
will read that into it.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>