RE: [dhcwg] I-D ACTION:draft-ietf-dhc-vendor-02.txt

Josh--

here's an excerpt from section 3:

   "This option contains information corresponding to one or
more
   Enterprise Numbers.  Multiple instances of this option
may be
   present, and MUST be concatenated in accordance with RFC
3396 [4].
   An Enterprise Number SHOULD only occur once among all
instances of
   this option.  Behavior is undefined if an Enterprise
Number occurs
   multiple times.  The information for each Enterprise
Number is
   treated independently, regardless or whether it occurs in
an option
   with other Enterprise Numbers, or in a separate option."

While I agree that it is very difficult to specify
appropriate behavior of the server if a client sends
[apparently] ambiguous data (i.e., multiple instances of an
Enterprise Number), I'm becoming less and less favorably
disposed towards protocol behavior that is undefined or
implementation-dependent.  I'd prefer to see something like
the following:

   "An Enterprise Number MUST only occur once among all
instances of this
   option.  If an Enterprise Number occurs more than once,
despite this
   restriction, a DHCPv4 server SHOULD process only the LAST
instance."

I suggest this wording because I believe that the MUST is
what you really intend in this section, and the second
proposed sentence specifies what would happen if a client
violates the restriction.

I'm open to suggestions and other opinions about this,
however.

--Barr

_______________________________________________
dhcwg mailing list
dhcwg@ietf.org
https://www1.ietf.org/mailman/listinfo/dhcwg