Date: Fri, 28 Sep 2001 08:45:10 -0400
From: Valdis(_dot_)Kletnieks(_at_)vt(_dot_)edu
Message-ID:
<200109281245(_dot_)f8SCjA310953(_at_)foo-bar-baz(_dot_)cc(_dot_)vt(_dot_)edu>
| I assume that what you're asking is whether for exceptions (1) and (2),
| if the intent is that an ICMPv6 error MAY, SHOULD, or MUST be generated?
|
| If so, then you are correct - the spec should probably read:
No, that's not needed. All that is needed is to refer to the sections
cited - 3.2 for "too big" says MUST, 3.4 for "parameter problem" says
SHOULD.
The reason that all that "except for ..." is just to avoid contradictions
in the document.
That is, one section says
if X happens you MUST do A.
another section says
if Y happens, you MUST NOT do A
then something has to resolve the situation when X and Y happen
simultaneously.
In this case "A" is send an ICMP, X is "packet is too big", and Y
is "packet is sent to a multicast address".
To fix the problem (which as worded above, in this message) would be
if X and Y happen you both MUST and MUST NOT do A
the Y rule is changed to
if Y happens, you MUST NOT do A, unless X happens as well.
Then it is clear which of the two rules is to apply when we have
both X and Y, and the contradiction is averted.
Nothing needs fixing here.
| Any packet warriors with more experience than I want to clarify this?
To discover the reasons why ICMPs are sent in these cases you can
go back and look in the archives - but I don't think that was the
actual point of the original question. More one of how to interpret
what to do in the exception cases. From the rule containing the
exception alone you cannot - but read that with the rest of the doc
and all is clear.
kre