Origins of DATE command

[markd at mira.net.au]

> > In the arguments over the NNTPv2 spec, one of the concerns was that
> > clients would use the DATE command as a remote time clock, and beat
> > the NNTP server to death asking about the time of day.  [In today's

> Let me be a polyanna and suggest that we simply return the date in
> YYYYMMDDHHMM format... that is, we don't return the number of seconds.

Perhaps YYYYMMDDHHMM00 would break less existing code?

However, with daytime ports and the likes is it really likely that a
wildly popular time setting client will go to an nntp port?

I'm sure many/all people on this list are aware that free/shareware
progs exist on windows platforms that use ntp/rdate so programmers
capable of building a better mouse-trap will hopefully be aware of the
competition.

Magic cookie is a good strategy to avoid all of this but it does open
up some interesting questions for replicated servers and, as Tom says:

> I'd prefer not to invent a lot in this area.  Let's spend our
> "invention" budget on search mechanisms, improved transports, and truth,
> justice, and 8-bit text.


Ahem. Another thought. Perhaps this should be the date/time of last
article written to the database rather than the current date. The
standard could be worded appropriately to protect existing servers
that return the current date "...no new articles have arrived since
the date returned by this command.". This would also protect nnrp-type
programs that determine the current database state periodically.

But it might open up more of a pandora's box if impatient newsreaders
disconnect and re-connect to force the nntp server to refresh its
state, or does that already occur - sigh.



Regards.