[NNTP] LIST EXTENSIONS (again)

[Andrew - Supernews]

>>>>> "Mark" == Mark Crispin <MRC at CAC.Washington.EDU> writes:

 Mark> Well, then, I guess that it comes to whether more harm is
 Mark> caused by having to change:

 Mark>   . a configuration (installed on a handful of sites at most)
 Mark>   of one server implementation,

 Mark>   . break until changed every client based upon the c-client
 Mark>   library (most notably Pine -- we're talking about at least
 Mark>   tens of thousands if not hundreds of thousands of separate
 Mark>   implementations -- as well as break the fixed version of
 Mark>   these clients against Diablo servers.

You don't seem to understand that this is _already broken_. None of
the MODE READER-related stuff in the spec is anything more than trying
to document _current practice_ in such a way as to get interoperability.
Trying to do STARTTLS or AUTHINFO before doing MODE READER (assuming a
reader client) does not interoperate with existing software in certain
known (but rare) cases and there is no workaround.

(Proper handling of 480 responses allows a client to cope
transparently with the INN vs. Diablo differences. Even if we mandate
the order, we should probably point this out so that new clients can
handle old noncompliant servers.)

 Mark> If it helps matters, I would agree to a separate
 Mark> AUTHINFO/STARTLS command for clients vs. peers in return for
 Mark> the deprecation of MODE READER (replacing it with a no-op and
 Mark> servers MUST NOT require).

You still don't seem to be getting the point here. If MODE READER could
be disposed of that easily, we'd have done it long ago. Nobody likes it;
it's a really nasty wart on the protocol.

-- 
Andrew, Supernews
http://www.supernews.com