ietf-nntp Section 11.5 - NEWNEWS

[Paul Overell]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

In article <14874.24383.652008.668188 at gargle.gargle.HOWL>, Lee Kindness
<lkindness at csl.co.uk> writes
>Paul Overell writes:
> > Whether from POSIX time or TAI or NTP or GPS or the speaking clock or a
> > video camera mounted on a sun dial :) - how a server obtains UTC is not
> > our problem.
> > The DATE command returns UTC.  It SHOULD be accurate.
>
>Why? What in the spec requires an accurate time? I believe the DATE
>command should return an arbitrary string/token - The client shouldn't
>care in the slightest what it parses to - DATE is only really being
>used to mark/note the time of the last NEWNEWS (or NEWGROUPS) command
>issued. Clients should file this away and then resurrect it (and parse 
>it slightly) for sending as an argument to the next NEWNEWS/NEWGROUPS
>call...
>
> [c] DATE
> [s] 111 20001121113200
> [c] (stores date for future use, sends NEWGROUPS command with last
>     saved date, or current (client) date)
> [c] QUIT
>
>And on the next connection:
>
> [c] DATE
> [s] 111 20001122120000
> [c] NEWGROUPS 20001121 113200 GMT
> [s] 230 list of new newsgroups follows 
> [s] .
> [c] QUIT
>

Requiring NEWNEWS and NEWGROUPS to use the "arbitrary string/token" from
the DATE command is an incompatible change from RFC977.

RFC977 does not have the DATE command so RFC977 conforming clients have
to obtain UTC from their own sources.

The only way to be compatible with RFC977 is if NEWNEWS and NEWGROUPS
take UTC, therefore DATE must return UTC.

>The DATE command MAY be accurate.

You can't mean this - using MAY would imply that a server could *always*
return

111 11111111111111

and still claim to be compliant!

>
>I also think all references to UTC as command arguments should be
>removed from the spec - this is not current practise and does not buy
>us anything over just using GMT as the token. GMT and UTC are being
>used interchangeably in the specification, so one token should
>suffice.
>
>Lee.

For compatibility with RFC977 we are stuck with the token "GMT" to mean
UTC for the syntax of the NEWNEWS and NEWGROUPS commands.  One token
does suffice.

However, when specifying semantics we should be precise about what we
mean.  The timescale used in the DATE, NEWNEWS and NEWGROUPS command is
UTC.


Regards
- -- 
Paul Overell                                             T U R N P I K E

-----BEGIN PGP SIGNATURE-----
Version: PGPsdk version 1.7.1

iQA/AwUBOhpwnB8WRy+wdge+EQJYiwCg964qfMSSog7MU9zoA3uI8/vkpzQAn352
8i+q4eLYDN6G3626S7v/C3XX
=0K04
-----END PGP SIGNATURE-----