RE: opsec and 2119 keywords

[George Jones <gmj@pobox.com>]

OK.  One more time on 2119.    It looks like the options are

  1. Leave UPPERCASE keywords in per 2119 as the are now.
     There is no apparent technical conflict with using them this way in an
     info, but there are people with reservations about whether we
     SHOULD (should ?) do this in a doc that is non-BCP.

  2. 1,$s/MUST/must/g
     1,$s/SHOULD/should/g...

     This would address the concerns, allow the doc to get out quickly
     as info (per consensus of the BoF), but weaken the strength of
     the reqs.  Would probably also add some verbiage making it explicit
     that these "musts", "shoulds" etc DO NOT have the force of 2119
     and that individual operators will have to decide for themselves
     which reqs are MUSTs for their situation.

  3. Do something more "drastic", like restructuring entire doc into
     smaller chunks per Fred Baker's suggestion or putting the MUSTs
     info implementation section per Pekka's suggestion (but would
     "MUST implement (RADIUS|DIAMATER|TACACS+) violate 2119 sec. 6 ?).

It seems to me that 2 is going to be the best path to meeting the
goals of getting the ideas out in a usable form without
bending/violating conventions.   A working group, if formed,
seems to be the best place to address 1 and 3.

If I hear no hue-and-cry to the contrary, this is the way I will
proceed.

Thanks,
George M. Jones    |  Omne ignotum pro magnifico est ("Everything unknown is
                   |  taken as marvelous")
                   |
                   |      Tacitus (Agricola)