Re: RFC 1812 Clarification

[Fred Baker <[e-mail]>]

At 3:31 PM 9/15/95, Mike Benefield wrote:
>We are going through 1812 with a fine-tooth comb in order to develop our
>MIL-STD PICS and when we're finished we would like to share our 1812 notes
>with you for whatever value they may be in support of your editorial work.

Thank you

>In the meantime, we have just two short issues that we'd like to share with
>you and we ask for your thoughts if you have a moment.
>
>     1.  IGMP Issue:
>
>          In paragraph 4.2.3.2 (on top of page 51), we read (for local IP
>multicasting), "This implies support for all of the RFC 1112 including IGMP
>(see section [4.4])."  Now, in paragraph 4.4, "a router SHOULD implement the
>host part of IGMP."
>
>               This 4.4 "mandatory" (SHOULD) requirement is only a part of
>RFC 1112 while 4.2.3.2 "implies" support for "all" of RFC 1112 (including
>the IGMP itself).
>
>          IGMP used to be optional for local multicasting and now it seems
>to be mandatory according to the "implied" support in 4.2.3.2 above.
>
>          Our two questions are:
>
>               Should the "implies" in 4.2.3.2 be a "SHOULD" or "MUST"?
>
>               Should IGMP be mandatory for our router multicasting
>requirements profile or are we  covered with just implementing the "host
>part" of IGMP (and we could use a bit of an indication about the definition
>of the "host part")?


"MUST" means "Your Router is Incorrect and broken if it doesn't do this".
"SHOULD" means "This is important, but either cases exist where it isn't
required or valid alternatives exist."

Not all routers implement Network IP Multicast (PIM, DVMRP, etc).
Therefore, we cannot say "You MUST Implement IP Multicast Features".
However, if you DO implement IP Network Multicast, you must implement
certain things, and IGMP is one of them.

This is different than local multicasting. IGMP is not used for local
multicasting, such as OSPF neighbor to neighbor multicasts. IGMP is used by
hosts and potentially by routers to join network layer IP Multicast groups.


>     2.  ICMP issue:
>
>          In paragraph 4.3, most of the material through 4.3.3.3 seems to be
>taken from RFC 1122 in some fashion (or is similar), but without reference
>to RFC 1122 paragraphs.  Are these 1812 paragraphs replacing 1122 paragraphs
>or are they simply modifying them?
>
>          Paragraphs 4.3.3.4 thorough 4.3.3.9 basically reference RFC 1122
>in a modifying/supportive sense.
>
>          It's as if someone started through paragraph 4.3 with the aim in
>mind of replacing 1122, but stopped at 4.3.3.4.
>
>          What we really need to know is how much of 1122 has been replaced
>by paragraph 4.3 and consequently what remains in RFC 1122, paragraph 3.2.2,
>for successful router ICMP implementation.  Or, are all of paragraph 4.3
>subparagraphs simply modifications to 1122 requirements?
>
>          One last thought:
>
>               RFC 1812 paragraph 4.2.1, IP, tells us what is happening with
>RFC 1122 paragraph 3.2.1 right up front.  Should there be an "RFC 1122,
>paragraph 3.2.2 compliance" statement for ICMP in RFC 1812 paragraph 4.3.1?


4.3 is a modification of 1122, but as you say is fairly thorough; routers
do more with ICMP than hosts do. Your suggestion is a good one, and when it
comes time to revise this I will act on it.

For a router, the router should comply with 4.3; I believe that it contains
enough information that you need not reference 1122.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
So Bill Gates has paid Mick and the boys 12 million dollars to use "Start Me
Up" to sell Windows 95.

For those who don't know the song, it's the one that goes "You make a grown
man cry...".

Food for thought