Re: draft-ietf-dhc-dhcpv6-opt-timeconfig-02.txt

[Bill Fenner <fenner@research.att.com>]

Examples i. and ii. are valid settings for the TZ environment
variable according to POSIX 1003.1-2001 / SUSv3 -- so it's not that
they're inventing a format, just reusing [and restating] one that was
standardized elsewhere.  (Although they skipped the bit about quoting
the std or dst fields if they contain + or - characters, which we might
want them to include if we think using POSIX format is OK.)  I don't
like option iii. that much because it is not portable (e.g. a device
that shows up on the network that doesn't have the database to map
"America/New_York" to something useful is helpless) or it could mean
different things for different devices (e.g. Australia changes its
timezone rules, and some devices have the 2000 version of the database
and some have the 2003 version).

(On the other hand, on most systems that use the Olson time zone database,
the filename method (e.g. "America/New_York") is the only way to set
the time zone system-wide; setting a TZ environment variable for all
processes could be difficult or impossible.  What a hard problem they're
trying to solve =( )

iCalendar is the only IETF standard I'm aware of that has a way to
represent timezones and DST shifts, but it is way too verbose for
DHCP.

  Bill