Randy and Alvaro: 

A few point have raised my disagreement regarding the AD reviews.   At this
point, the text is at odds with RFC4271.   If you feel RFC4271, does not
speak about these issues - you may have missed why the BGP FSM is there.
Please change your text. 

Sue  

Point 1 - 
===============
Alvaro: 
>   An implementation that advertises support for BGP Extended Messages
> MUST be capable of receiving an UPDATE message with a length up to
 > and including 65535 octets.

Randy: 
>like many MUSTs, there is no direct enforcement, only the ability to
diagnose blame in the case of failure and have the resulting tac ticket be
classified as an enhancement request :)

Sue:  Normally, the diagnosis comes from testers that people run prior to
putting things in the field (E.g. IXIA).   Part of BGP implementation
testing is a length BGP FSM test with a tester.  
-------------
Point 2:   Error handling is precisely defined in the BGP FSM.  You need to
tag the Extended message reception to the BGP Header error for all of this
to work. 

Randy's comment: 
Randy: "ok, jgs, what did you mean by "any speaker that treats an improper
extended message as a fatal error," and what did you think we should do
about it.  neither 4271 not 7606 tell what to do with a fatal error per se."


Sue's response:   
Any messages that exceeds 4096 for BGP speaker will result in a BGP state
machine event of BGP Header error (event 21,  BGP state machine, page 49).
Therefore,  the error procedures would be based on this even in the BGP FSM
based on state.    The RFC4271 error handling procedures are defined by the
BGP FSM based on the current state .   Please walk through the BGP FSM to
fine the precise action based on state.  Implementations do adhere to it.
There are options - so please read the entire section. 

Text from version 21

   A BGP speaker that has the ability to use Extended Messages but has
   not advertised the BGP Extended Messages capability, presumably due
   to configuration, SHOULD NOT accept an Extended Message.  A speaker
   MAY implement a more liberal policy and accept Extended Messages,
   even from a peer to which it has not advertised the capability, in
   the interest of preserving the BGP session if at all possible.

   A BGP speaker that does not advertise the BGP Extended Messages
   capability might also genuinely not support Extended Messages.  Such
   a speaker would be expected to follow the error handling procedures
   of [RFC4271] if it receives an Extended Message.  Similarly, any
   speaker that treats an improper Extended Message as a fatal error,
   MUST treat it similarly. 

   The inconsistency between the local and remote BGP speakers MUST be
   flagged to the network operator through standard operational
   interfaces.  The information should include the NLRI and as much
   relevant information as reasonably possible.

Replacement text.  

    A BGP speaker that has the ability to use Extended Messages but has
   not advertised the BGP Extended Messages capability, presumably due
   to configuration, SHOULD NOT accept an Extended Message.  
   A BGP speaker that does not advertise the BGP Extended Messages
   capability might also genuinely not support Extended Messages.  
   If the BGP speaker does not accept an extended messages, the
   BGP speaker MUST indicate a BGP Header error (event 21) to the 
   BGP finite state machine (FSM).   

   A speaker MAY implement a more liberal policy and accept Extended
Messages,
   even from a peer to which it has not advertised the capability, in
   the interest of preserving the BGP session if at all possible.

 
============================================================================
=========================

#3  - on reporting this error - The HEADER length error is reported as part
of the BGP Statement machine for errors. 

Why are we going through additional issues rather than stay with the BGP
Finite state machine.  Either the Length field is correct, or it is not.
If you have an erroneous length for what is configured, you drop it.   Most
people include the erroneous header length and check the BGP speakers.   Why
are you making this more complicated? 
 
If you want to enforce the MUST, then we must open the BGP FSM.   

========================
#4 - on the limit the message size, this was to be a warning to BESS or
other groups to not defined BGP applications which not stuff more junk in
than the BGP packet size can hold. 

> P4. I don't understand what this means: "Applications generating 
> messages which might be encapsulated within BGP messages MUST limit 
> the size of their payload to take into account the maximum message 
> size.

It is a social warning, like the cancer warning on cigarettes.   I not in
favor of social warnings in a base technical specification.   Author/AD can
remove. 


Sue Hares 

-----Original Message-----
From: Randy Bush [mailto:randy@psg.com] 
Sent: Sunday, March 5, 2017 2:47 AM
To: Alvaro Retana (aretana)
Cc: draft-ietf-idr-bgp-extended-messages@ietf.org; idr-chairs@ietf.org;
idr@ietf.org; Susan Hares
Subject: Re: AD Review of draft-ietf-idr-bgp-extended-messages-20

quite the review.  decided to earn your dinner, eh?  owe you one.  you can
collect in chicago.

> M1. Section 4. (Operation): "An implementation that supports the BGP 
> Extended Messages MUST be prepared to receive an UPDATE message that 
> is larger than 4096 bytes." Only UPDATEs?

yes.  see jeff's message.

> I know that the most likely case for exceeding the 4k size is an 
> UPDATE, but why are the other messages not considered?

rather ask why they should be.  i know this is bgp, but do we really need to
complicate life?

and there is a bit of a recursion problem if you try to extended OPEN.

> what does "prepared to receive" mean, and how can "MUST be prepared to 
> receive" be enforced?

   An implementation that advertises support for BGP Extended Messages
   MUST be capable of receiving an UPDATE message with a length up to
   and including 65535 octets.

like many MUSTs, there is no direct enforcement, only the ability to
diagnose blame in the case of failure and have the resulting tac ticket be
classified as an enhancement request :)

> Given the discussion in Section 5 (Error Handling), you might want to 
> add something like ".even if the Capability is not advertised".

this section is predicated on advertisement.  personally, as a long term
naggumite (yes, dr postel and i argued about this), i think the whole
send/receive without advertisement is a useless slippery slope and i would
be happy to toss it.

> M2. Section 5 (Error Handling).  "A BGP speaker that has the ability 
> to use extended messages but has not advertised the BGP Extended 
> Messages capability, presumably due to configuration, SHOULD NOT 
> accept an extended message.  A speaker MAY implement a more liberal 
> policy and accept extended messages even from a peer that has not 
> advertised the capability."  This paragraph troubles me a lot because 
> it is in direct contraction with Section 3: "A peer which does not 
> advertise this capability MUST NOT send BGP Extended Messages, and BGP 
> Extended Messages MUST NOT be sent to it.".  However, I think that 
> John Scudder's reasoning [3] makes sense ("keep the session up at
> (almost) all costs", and there's clear precedence in the WG) for the 
> case where the sender did advertise the Capability, but I'm not 
> convinced on the case where it didn't - please include something like 
> John's explanation in the text.

you did see where john went on to say "I acknowledge this choice would be
fairly disgusting, and I understand why the authors didn't make it."

but, if we keep the MAY, then i guess throwing in jgs's excuse is
worthwhile.

   A BGP speaker that has the ability to use extended messages but has
   not advertised the BGP Extended Messages capability, presumably due
   to configuration, SHOULD NOT accept an extended message.  A speaker
   MAY implement a more liberal policy and accept extended messages,
   even from a peer to which it has not advertised the capability, in
   the interest of preserving the BGP session if at all possibe.

> M2.1.  Even with John's explanation, I find myself thinking that this 
> specification could result in some sloppy implementations: if I need 
> to account for receiving unexpected Extended Messages in my code, then 
> maybe I won't worry too much about controlling what to send to my 
> peers - specially in cases where it would be easy to just replicate an 
> UPDATE (like in a peer-group) and not worry about possible exceptions.
> I know that we can't avoid bad implementations, no matter what text is 
> added - but I think that recognizing the threat (maybe in the Security 
> Considerations section) of someone receiving an Extended Message when 
> they don't support it would be good.  I know that there's text in the 
> document already which talks about what to do if the receiver doesn't 
> support Extended Messages - I'm just worried about potential issues 
> with memory allocation if the receiver was not ready.

in the absense of an AD with the guts to let me tear that out, i'll put a
note in sec cons. :)

   Section 5 allowed a receiver to accept an extended message even
   though they had not advertised the capability.  This slippery slope
   will surely lead to sloppy implementations sending extended messages
   when the receiver is not prepared to deal with them, e.g. to peer
   groups.  At best, this will result in erroes; at worst, buffer
   overflows.

> M3. Section 5 (Error Handling).  ".reset the session with a Bad 
> Message Length NOTIFICATION." Please be clear and specific: (something 
> like this would be more precise) "...send a NOTIFICATION message with 
> the Error Code set to Message Header Error and the Error Subcode set 
> to Bad Message Type, and close the session".  Alternatively, you can 
> just remove the text (after the reference to rfc4271) as that is what
> rfc4271 already says and there's no need to repeat it here and risk 
> not being precise.

i like removing text!

> M4. Section 5 (Error Handling).  "Similarly, any speaker that treats 
> an improper extended message as a fatal error, MUST do likewise."  It 
> sounds that you're saying that any fatal error will result in a "Bad 
> Message Length NOTIFICATION".  I hope that is not what you meant - and 
> that other errors should result in the appropriate action from 
> rfc4271/rfc7606.  IOW, the error checking for the message (besides de
> length) doesn't change, right?

ok, jgs, what did you mean by "any speaker that treats an improper extended
message as a fatal error," and what did you think we should do about it.
neither 4271 not 7606 tell what to do with a fatal error per se.

saying treat it as an as an UPDATE Message Error per RFC4271 is not perfect,
as it could be an overly long message of a different type.

clue bat, please.

> M5. Section 5 (Error Handling).  "The inconsistency between the local 
> and remote BGP speakers MUST be reported via syslog and/or SNMP."
> SNMP?  AFAIK, there's no object that can report this inconsistency 
> since there's no NOTIFICATION generated.  In the proposed text by 
> Gunter Van De Velde [4], SNMP and syslog were mentioned as examples - 
> I suggest you follow that path (no need for all the "flowery
> language") and just reference mechanisms by example to avoid having to 
> point at how it would be done.
>
> M5.1. [minor] Gunter had originally suggested that the message that 
> caused the inconsistency be included in the report.  Are you expecting 
> the inconsistency report to just be a "Bob sent me an Extended 
> Message, but I didn't advertise the Capability to him"-type message, 
> or something more?  It might be useful to provide some guidance 
> indicating what type of information might be useful/interesting.

   The inconsistency between the local and remote BGP speakers MUST be
   flagged to the network operator through standard operational
   interfaces.  The information should include the NLRI and as much
   relevant information as reasonably possible.

> M6. Updates to rfc4271.  The Abstract/Introduction correctly mention 
> that this document Updates rfc4271.  But I think we need to be more 
> specific, specially where rfc4271 changes and there is normative 
> language involved.  I found two cases:
> 
> M6.1. Section 5 doesn't describe the behavior if the message is longer 
> than 65535.  Please include either an explicit update to 
> rfc4271/Section 6.1 for the use of Extended Messages, or the specific 
> process here.
> 
> M6.2. rfc4271: "The value of the Length field MUST always be at least
> 19 and no greater than 4096" That needs to be updated to 65535.

6.  Changes to RFC4271

   [RFC4271] states "The value of the Length field MUST always be at
   least > 19 and no greater than 4096."  This document changes the
   latter number to 65535 for UPDATE messages.

   [RFC4271] Sec 6.1, specifies raising an error if the length of a
   message is over 4096 octets.  For UPDATE messages, iff the receiver
   has advertised the capability to receive extended messages, this
   document raises that limit to 65535.

> M7. What about transition/migration/partial deployment?  What should 
> the behavior be if, for example, an Extended Message UPDATE is 
> received from a peer, but can't be propagated to others because they 
> don't support Extended Messages

if they do not support extended messages, then they can not be bgpsec
speakers.  so the whole bgpsec path stripping applies and the message
becomes short.

> There should be some guidance for the general case (i.e. when the 
> total size is >4k due simply to the total amount of information, and 
> not because a single attribute, for example, is really big)

how about "do not do this?"

i believe you have entered a twisty maze in which all rooms do not have
path(s) to the exit.

4.  Operation
...
   A BGP announcement will, in the normal case, propagate throughout the
   BGP speaking Internet; and there will undoubtedly be BGP speakers
   which do not have the Extended Message capability.  Therefore putting
   an attribute which can not be decomposed to 4096 octets or less in an
   Extended Message is a sure path to routing failure.

> P1. Abstract: s/extend its current message size from 4096/extend its 
> current maximum message size from 4096

ack

> P2. s/I-D.ietf-sidr-bgpsec-overview/I-D.ietf-sidr-bgpsec-protocol

ack

> P3. IANA already assigned Code 6 for the Capability.  Please use that 
> value and remind IANA of the early allocation in the IANA 
> Considerations section.

ack

> P4. I don't understand what this means: "Applications generating 
> messages which might be encapsulated within BGP messages MUST limit 
> the size of their payload to take into account the maximum message 
> size."

given that we restrict to UPDATE, i do not know what this means at all
:)

> P5. Security Considerations: I think it would be good to also 
> reference rfc4272 (BGP Security Vulnerabilities Analysis) in this 
> section.

sure

> N1. "It does enable large BGPsec BGPSEC_PATHs, see 
> [I-D.ietf-sidr-bgpsec-protocol]."  Nice, but superfluous as it refers 
> to this document.  [2 <text/html; utf-8 (base64)>]

ack

randy, who is serious about the dinner offer.  this was one hell of a
       review