I take Sharon's points but I haven't quite convinced myself
they add up to a DISCUSS. So I will log this review as a COMMENT
and would be interested to see some conversation about it.
(There is a minor bureaucratic DISCUSS issue, however, which
I will document shortly.)
Brian
On 2007-01-18 21:26, Sharon Chisholm wrote:
I have been selected as the General Area Review Team (Gen-ART) reviewer
for this draft (for background on Gen-ART, please see
<http://www.alvestrand.no/ietf/gen/art/gen-art-FAQ.html>
Please resolve these comments along with any other Last Call comments
you may receive.
Document: Modifying ICMP to Support Multi-part Messages
Summary: This document is on the right track, but there are open issues
that should be resolved before publishing
Comments
---------
1. I don't see why more care wasn't taken to ensure interoperability
with existing ICMP specifications and implementations.
1.1 Section 5 should really define three classes of applications
- (legacy) support non-extended
- (legacy support non-extended and non-compatible extensions
- (new) support non-extended, non-compatible extensions and
compatible extensions
It should list how a consumer can tell the messages apart and what
implementers can do to maximize interoperability.
1.2 Each of the sections in section 5 should be more clear as to whether
or not backwards compatibility (interoperability) is guaranteed,
possible or not possible. It reads more like the results of a lab
experiment then requirements against implementations. Reading between
the lines it seems like section 5.1 can't guarantee compatibility but it
is likely unless implementers do some of the things noted. Similarly for
5.2. 5.3 seems a bit more promising.
1.3 Section 5.5 should make it mandatory to "include a non-default
operation mode to also interpret non-compliant responses."
Nits
-----
1. Shouldn't the title of the document be more along the lines of
"Extended ICMP to Support Multi-part Messages" or "Modified ICMP to
Support Multi-part Messages"
2. In section 2, last paragraph, last sentence, it says "whenever a
packet is dropped to to TTL expiration."
3. In section 3, it says "The following is a summary of changes to ICMP
that are proposed by this memo:", but that should likely now read
"introduced by this memo".
4. Section 4 lists all the ICMP messages and says that "Many ICMP
messages are extensible as currently defined" but doesn't say which are
which. This information is given in section 4.6. It would seem cleaner
if these were combined into a single section.
5. Section 4, after the bullets, the first paragraph starts with "Many
ICMP messages are extensible as currently defined" while the second
paragraph starts with "Many ICMP messages are not extensible as
currently defined" which is awkward. I would suggest replacing one or
the other with "Some" or "Many other" or something to indicate which is
the more common or be more neutral. They both can't be many.
6. In section 4, on page 7, about half way down, it says "Because the
sixth octet of each of the impacted ICMPv4 messages was reserved for
future use" which explains how we can just start using these bits. This
explanation should come earlier in the document.
7. In section 4.6, last paragraph, it raises a requirement against
future ICMP extensions which seems a bit too buried in the document.
8. In section 5, third paragraph is says "Some ICMP implementations,
produced between 1999 and the present," but the present is a moving
target. Suggest specific date or say "time of publication of this memo".
9. In section 5.1, first paragraph, last sentence it says "To date, only
two applications falling into this category have been identified and the
degree to which they are effected is minimal." Is this saying they
mostly behaved as expected as defined in the ICMP documents? The phrase
is way too loose. Similarly the next paragraph talks about minimal
impact.
10. In section 5.2, fifth paragraph, it says "parse an ICMPv4 message
incorrectly " which is a bit ambiguous. Do we mean tradition ICMPv4
messages or the ones defined in this memo or for this use case do they
look the same?
11. In section 5.2, fifth paragraph, the bullets list three conditions
under which the message won't get parsed properly, two of which don't
seem to apply to the use case of 'Non-compliant Application Receives
ICMP Message With No Extensions'
12. In section 5.2, last paragraph is says "A similar analysis can be
performed for ICMPv6. However, the numeric
constants would change as appropriate." This should really read about
the expectation on achieving backwards compatibility (interoperability),
not the expectation on lab results.
13. In section 5.3, second paragraph it says "Provided that the entire
ICMP messages does not exceed." That should be "message"
14. In section 5.3, second paragraph is says "However, each vendor will
decide how many octets to include." That should say "However, each
implementation will decide how many octets to include."
15. Section 5.3 should be written in the context of how to achieve
interoperability/backwards compatibility.
17. In section 5.5, last paragraph it says "will treat the following
octets" but I think that should be "will treat the octets that follow"
to not confuse it with document text indicating that the list will be
outlined below in the RFC.
18. In section 6, first paragraph, last sentence says "by the first 128
octets of the "original datagram" field, ICMP extensions and NAT will
not interact with one another." Interact seems a poor choice of words.
It can be interpreted as either interoperate (good) or interfere (bad).
19. In section 6, second paragraph, is it saying you will get
interaction with NATS unless implementations do something that is not
explicitly forbidden or is it something that is illegal?
20. In section 7, first paragraph, it says "This memo proposes an
optional ICMP Extension Structure that can be appended to the ICMP
messages referenced in Section 4.6 of this document" which makes it
sounds like the messages were originally defined in this document, but
they weren't. An additional reference to the RFC(s) they were defined in
would help.
21. In section 7, second paragraph, it says "MAY process selected
objects while ignoring others." which makes it sound arbitrary. I
believe based on what is in the next sentence that they are only
ignoring stuff they don't understand?
22. In section 7, the version of this stuff is defined to be 2. Do other
ICMP messages have different versions? Is this a way to tell them apart
when they are received?
Sharon Chisholm
Nortel
Ottawa, Ontario
Canada
_______________________________________________
Gen-art mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/gen-art
_______________________________________________
Gen-art mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/gen-art