[ntp:hackers] "bulletin C via DNS" service on line

Warner Losh imp at bsdimp.com
Mon Nov 23 03:03:52 UTC 2015


On Sun, Nov 22, 2015 at 5:27 PM, Hal Murray <hmurray at megapathdsl.net> wrote:

>
> phk at phk.freebsd.dk said:
> > Because there is a good and widely available API for getting A records,
> but
> > TXT records are a nightmare to get hold of.
>
> Any reason not to have both?
>
> I can get TXT records via command line tools.  In scripting environments,
> it
> may be easier to process text strings than do the arithmetic to decode a
> dotted-quad.


TXT records are large enough you could publish the entire table in the 255
character limit. At 5 characters per leap, you can get 51 leaps in before
you run out of room. Using a 3 character per leap encoding, you could get
85 (assuming we don't go 100 years without one). 50 leaps into the future
is ~75 years (assuming people know the table as of Dec 1, 2015). 85 leaps
into the future is ~120 years. You could even do 2 characters per leap for
128 leaps good for ~200 years. Of course all these time frames exceed the
length of time we've been doing leaps...

5 character encoding: yyyyM. M is j or d for a + leap, J or D for a - leap
and o or O
to indicate no leap indicated by latest Bulletin C for june / december. 3
character
is yyM with the same encoding and an understanding that if a subsequent
entry
is smaller than a pervious one, it's a new century. 2 character would be
similar,
but would also need a rule for what to do if there's leaps in a 10 year
period.It's
expandable to add m and s, but those won't be needed in the next hundred
years.

The current leap with 2 character format would be:

1d2j2d3d4d5d6d7d8d9d1j2j3j5j7d9d0d2j3j4j5d7j8d5d8d2j5j5O

Warner


More information about the hackers mailing list