Re: [tcpm] WGLC comments addressed in draft-ietf-tcpm-converters-09?

[<[e-mail]>]

Hi Phil, 

Please see inline.

Cheers,
Med

> -----Message d'origine-----
> De : [e-mail] [mailto:[e-mail]]
> Envoyé : mercredi 21 août 2019 18:19
> À : BOUCADAIR Mohamed TGI/OLN; [e-mail];
> [e-mail]
> Cc : [e-mail]
> Objet : RE: WGLC comments addressed in draft-ietf-tcpm-converters-09?
> 
> Med,
> Coming back to this after hols...
> Thanks for the updates. In-line.
> phil
> 
> > -----Message d'origine-----
> > De : tcpm [mailto:[e-mail]] De la part de
> > [e-mail] Envoyé : jeudi 1 août 2019 10:58 À :
> > [e-mail]; [e-mail]; [e-mail]
> > Cc : [e-mail] Objet : Re: [tcpm] WGLC comments addressed
> > in draft-ietf-tcpm-converters- 09?
> >
> > Phil,
> >
> > I prepared an updated version with the following changes to address
> > your remaining comments:
> >
> > * Position Figure 1 right after the text about upstream/downstream
> > connections to avoid the confusion about the direction.
> 
> [phil] thanks, this certainly helps. It's a shame that the two bullets
> that actually define upstream / downstream connection are several pages
> later at the bottom of page 9 (can it be moved?).

[Med] The pointer to Figure 1 is sufficient IMO to understand what is meant by upstream/downstream connections. This is a matter of editorial taste.  

> One could also be pedantic about Figure 1 ("upstream" should be "upstream
> connection"; there's a missing space to the left of "Server" ;

[Med] Can be fixed. 

 the two
> interfaces are kind of below the Converter - Fig 1 seems to be a physical
> boxes picture rather than a protocol pic). I know what you mean and asci
> art is tricky, and maybe it's ok or the rfc editor will have suggestions.
> 
> > * Delete "(1)" from Figure 5 caption
> > * Add text to clarify why "eventually" is used in the text. Rearranged
> > the text about address preservation/sharing modes, accordingly.
> 
> I like that you moved the para about address preservation /sharing, so
> that it appears a bit earlier.
> The sentence about "eventually" is, for me, no clearer - sorry. Can't you
> just explain the two address modes one at a time?

[Med] The converter can behave in address preservation or address sharing modes: 
* The address sharing mode assumes that the converter uses a pool of IP addresses to service the clients. That is, the external IP address of packets relayed by the converter will belong to that pool. For incoming packets, the converter will need to systematically rewrite the destination IP address. 
* The address preservation mode assumes that the converter will use one of the client's IP addresses as source address when relaying packets. For incoming packets, the converter does not need to rewrite the destination address for the TCP case. For the MPTCP case, the converter will rewrite the destination address of incoming packets only if a subflow not bound to the preserved address is used to relay data.

These considerations are deployment-specific. A pointer to an external document is provided for further information.  

> 
> > * Section 3.2: remove the text about inserting Convert TLVs in
> > "subsequent messages".
> > * Section 3.3: add finally a side not to remind that RST does not
> > close an MPTCP connection, and hence is not reflected by the converter
> > on the TCP connection.
> 
> Thanks
> 
> > * Section 4 (Introduction): Clarified that both control and data
> > messages are sent over a relayed connection. I hesitated to add the
> > NEW text you proposed about relaying connections, but finally
> > discarded it because the behavior is already described in many places
> > in the documents. No need to be redundant.
> 
> I still think that the actual basic operation of the converter for data
> packets is weakly described (assuming the description is just in this
> document and not somewhere else)
> 
> S3.2 Theory of operation says " Any user data received by the Transport
> Converter over the upstream (or downstream) connection is relayed over the
> downstream (or upstream) connection." This is fine as a high level
> summary.

[Med] This is all what the converter has to do with user data.  

> S4 is really about the control protocol messages, not the data plane. But
> it does say: " By default, the Transport Converter listens on TCP port
> number TBA  for Convert messages from Clients.  Clients send packets bound
> to connections eligible to the conversion service to the provisioned
> Transport Converter using TBA as destination port number.  This applies
> for both control and data messages."
> As far as I can see, you don't actually say what the converter does with
> the data packets. Perhaps it is obvious, but it surely should be stated.

[Med] Already stated in Section 3. We don't have any additional behavior to specify.  

> The behaviour for the other direction should be stated.

[Med] Already covered in section 3. Section 4 is about the description of Convert messages (which apply for both directions).   

 And some statement
> about the behaviour in the address preservation /sharing modes. 

[Med] How address sharing/preservation is implemented is out of scope. That's said, the document includes a key requirement when address sharing is used (Section 3).

Also, what
> is done (MUST /SHOULD /MAY) if there's no 'conversion entry' for this
> client. Perhaps this can be partially done with a reference to a proxy RFC
> (presumably that would be an extra Normative ref).

[Med] Do we really need to say that packets are discarded when no entry is found? Especially, that the text in Section 3 says the following:

"If the check is successful, ..." 

And 

"  Then, when the Converter receives an incoming SYN, it checks its
   mapping table to verify if there is an active mapping matching the
   destination IP address and destination port of that SYN.  If an entry
   is found, ..."

> On minor points, I'd describe the data plane operation in its own sub-
> section. 

[Med] I don't think this is needed. Section 3 does already include the required details. 

I'd say "data" rather than "data messages".

[Med] Agree. 

> 
> > * Update Figures 11/19
> > * Add an appendix to record the design considerations from the changes
> > log.
> 
> Thanks
> phil
> 
> >
> > I also made some other edits to fix some nits.
> >
> > You may check the full diff at:
> > https://www.ietf.org/rfcdiff?url1=draft-
> > ietf-tcpm-converters-09&url2=draft-ietf-tcpm-converters-10
> >
> > Thank you for the review.
> >
> > Cheers,
> > Med
> >
> 
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]]
> Sent: 01 August 2019 09:58
> To: Eardley,PL,Philip,TUD1 R <[e-mail]>; Michael.Scharf@hs-
> esslingen.de; [e-mail]
> Cc: [e-mail]
> Subject: RE: WGLC comments addressed in draft-ietf-tcpm-converters-09?
> 
> Phil,
> 
> I prepared an updated version with the following changes to address your
> remaining comments:
> 
> * Position Figure 1 right after the text about upstream/downstream
> connections to avoid the confusion about the direction.
> * Delete "(1)" from Figure 5 caption
> * Add text to clarify why "eventually" is used in the text. Rearranged the
> text about address preservation/sharing modes, accordingly.
> * Section 3.2: remove the text about inserting Convert TLVs in "subsequent
> messages".
> * Section 3.3: add finally a side not to remind that RST does not close an
> MPTCP connection, and hence is not reflected by the converter on the TCP
> connection.
> * Section 4 (Introduction): Clarified that both control and data messages
> are sent over a relayed connection. I hesitated to add the NEW text you
> proposed about relaying connections, but finally discarded it because the
> behavior is already described in many places in the documents. No need to
> be redundant.
> * Update Figures 11/19
> * Add an appendix to record the design considerations from the changes
> log.
> 
> I also made some other edits to fix some nits.
> 
> You may check the full diff at: https://www.ietf.org/rfcdiff?url1=draft-
> ietf-tcpm-converters-09&url2=draft-ietf-tcpm-converters-10
> 
> Thank you for the review.
> 
> Cheers,
> Med
> 
> > -----Message d'origine-----
> > De : tcpm [mailto:[e-mail]] De la part de
> > [e-mail] Envoyé : mercredi 31 juillet 2019 14:22 À
> > : [e-mail]; [e-mail]; [e-mail]
> > Cc : [e-mail] Objet : Re: [tcpm] WGLC comments addressed
> > in draft-ietf-tcpm-converters- 09?
> >
> > Hi Phil,
> >
> > Thank you for double checking.
> >
> > Please see inline.
> >
> > Cheers,
> > Med
> >
> > > -----Message d'origine-----
> > > De : tcpm [mailto:[e-mail]] De la part de
> > > [e-mail] Envoyé : mercredi 31 juillet 2019 12:26 À :
> > > [e-mail]; [e-mail] Cc :
> > > [e-mail] Objet : Re: [tcpm] WGLC comments addressed in
> > > draft-ietf-tcpm-
> > converters-
> > > 09?
> > >
> > > I think most of my comments are addressed. Here are some things I
> > > think could still be clarified, plus a couple of extra questions
> > > that occurred to me when I was checking the latest version.
> > >
> > > Section 3.1
> > > <<Nevertheless, and unless this is explicitly stated,  the
> > > description assumes outgoing connections as default.>>
> > >
> > > This sentence seems to contradict itself (can something be both
> > > assumed and have to be explicitly stated?).
> >
> > [Med] Yes. Consider for example the following text:
> >
> >    "By default, the Transport Converter listens on TCP port number TBA
> >    for Convert protocol (Convert, for short) messages from Clients.
> >
> >    Clients send packets that are eligible to the conversion service to
> >    the provisioned Transport Converter using TBA as destination port
> >    number.  Additional information is supplied by Clients to the
> >    Transport Converter by means of Convert messages as detailed in the
> >    following sub-sections."
> >
> > It applies only for the outgoing connections.
> >
> >  Maybe:-
> > > In general this document assumes that the client initiates the
> > connection
> > > (in other words, it is an outgoing connection); the scenario with an
> > > incoming connection is discussed in a couple of places [references].
> >
> > [Med] I can use this wording if you think it is better.
> >
> > >
> > > In Figure 1 I find the 'upstream' and 'downstream' labels a bit
> > confusing
> > > (especially as the lines have arrowheads in both directions), and it
> > > is shown as the link between client and converter etc. I think it
> > > would be better to move lower down (ie separate from the actual
> > > link), something
> > > like:
> > > -------> upstream direction (outgoing connections)
> > > <------ downstream direction (incoming connections)
> > >
> >
> > [Med] Actually, upsteram and downstream are defined as follows:
> >
> >    o  the upstream connection is the one between the Client and the
> >       Transport Converter.
> >
> >    o  the downstream connection is between the Transport Converter and
> >       the Server.
> >
> > This is independent of the connection direction.
> >
> > > Figure 5 caption has a stray "(1)" that can be deleted
> >
> > [Med] Fixed.
> >
> > >
> > > Above Figure 6
> > > <<addresses and, eventually, the destination IP address and port
> number"
> > > I think ", eventually," should be deleted.
> > >
> > >
> >
> > [Med] "eventually" is justified: cover the case of a converter
> > configured in an address preservation mode (e.g., IPv6). The
> > destination IP address won't be rewritten in such case.
> >
> >
> > > Section 3.2 / 3.3
> > > There are two paragraphs at the end of 3.2 and a bit more in 3.3
> > > discussing what happens when a connection ends with FIN and TCP RST
> etc.
> > I
> > > think you should write a bit more about the MPTCP case - since there
> > > are subflow TCP RST and MP_FASTCLOSE cases to consider. A TCP RST on
> > > one
> > MPTCP
> > > subflow presumably shouldn't trigger the Converter to close the TCP
> > > connection on its other interface.
> >
> > [Med] Section 3.2 covers the generic TCP case. Hence, there is no need
> > to discuss MPTCP specifics in that section.
> >
> > I guess you are referring to this text in Section 3.3:
> >
> >    Note that, if the TCP connection fails for some reason, the Converter
> >    tears down the Multipath TCP connection by transmitting a
> >    MP_FASTCLOSE.  Likewise, if the Multipath TCP connection ends with
> >    the transmission of DATA_FINs, the Converter terminates the TCP
> >    connection by using FIN segments.
> >
> > The text covers exclusively the cases that lead to the termination of
> > the upstream/downstream connection.
> >
> > Given that MPTCP spec says:
> >
> >    "With MPTCP, the RST only has the scope of the
> >    subflow and will only close the concerned subflow but not affect the
> >    remaining subflows.  MPTCP's connection will stay alive at the data
> >    level, in order to permit break-before-make handover between
> >    subflows."
> >
> > the subflow RST is not covered (as it does not terminate the MPTCP leg).
> >
> > >
> > > Section 4 intro
> > >
> > > << This section describes the messages that are exchanged between a
> > >    Client and a Transport Converter.
> > >
> > >    By default, the Transport Converter listens on TCP port number TBA
> > >    for Convert protocol (Convert, for short) messages from Clients.
> > >
> > >    Clients send packets that are eligible to the conversion service to
> > >    the provisioned Transport Converter using TBA as destination port
> > >    number.  Additional information is supplied by Clients to the
> > >    Transport Converter by means of Convert messages as detailed in the
> > >    following sub-sections.
> > >
> > >    Convert messages may appear only in a SYN, SYN+ACK, or ACK.
> > >
> > >    Convert messages MUST be included as the first bytes of the
> > >    bytestream.  A Convert message starts with a 32 bits long fixed
> > >    header (Section 4.1) followed by one or more Convert TLVs (Type,
> > >    Length, Value) (Section 4.2).
> > > >>
> > >
> > > Some comments:
> > > The Client also listens on TCP port TBA (not just the converter)
> >
> > [Med] The client will listen on the internal port number that it
> > indicated when creating a mapping in the converter to allow for
> > incoming connections. This is needed to demux services hosted on the
> same client.
> >
> > This is covered in this text:
> >
> >    The Converter accepts the request by creating a TCP
> >    mapping (internal IP address, internal port number, external IP
> >    address, external port number).  The external IP address and external
> >    port number will be then advertised using an out-of-band mechanism so
> >    that remote hosts can initiate TCP connections to the Client via the
> >    Converter.  Note that the external and internal information may be
> >    the same.
> >
> > > Stress that ALL convert msgs start with the same header.
> > > I think the "Clients send packets..." para is better re-arranged.
> > >
> > > Question: there seems to be a contradiction. The text here says
> > > "Convert messages may appear only in syn, syn-ack, ack". But then in
> > > S3.2 it says "This information is sent at the beginning of the
> > > bytestream, either directly in the SYN+ACK or in a subsequent
> > > packet." (this information is "about the TCP options that were
> > > negotiated with the Server.") (Incidentally, in S3.2 essentially the
> > > same sentence is repeated two sentences later.)  is the idea that SYN
> / syn-ack /ack is the 'normal'
> > > case, but can be in later pkts?
> >
> > [Med] Good catch.
> >
> > OLD:
> >    The Client sends a SYN destined to the Transport Converter.  The
> >    payload of this SYN contains the address and port number of the
> >    Server.  The Transport Converter does not reply immediately to this
> >    SYN.  It first tries to create a TCP connection towards the target
> >    Server.  If this upstream connection succeeds, the Transport
> >    Converter confirms the establishment of the connection to the Client
> >    by returning a SYN+ACK and the first bytes of the bytestream contain
> >    information about the TCP options that were negotiated with the
> >    Server.  This information is sent at the beginning of the bytestream,
> >    either directly in the SYN+ACK or in a subsequent packet.  For
> >    graphical reasons, the figures in this section show that the
> >    Transport Converter returns this information in the SYN+ACK packet.
> >    An implementation could also place this information in a packet that
> >    it sent shortly after the SYN+ACK.
> >
> > NEW:
> >    The Client sends a SYN destined to the Transport Converter.  The
> >    payload of this SYN contains the address and port number of the
> >    Server.  The Transport Converter does not reply immediately to this
> >    SYN.  It first tries to create a TCP connection towards the target
> >    Server.  If this upstream connection succeeds, the Transport
> >    Converter confirms the establishment of the connection to the Client
> >    by returning a SYN+ACK and the first bytes of the bytestream contain
> >    information about the TCP options that were negotiated with the
> >    Server.
> >
> >
> > >
> > > Question: the text says "Clients send packets that are eligible to
> > > the conversion service to the provisioned Transport Converter using
> > > TBA as destination port number." Is this referring to the exchange
> > > of Convert protocol messages? Or is this referring to subsequent
> > > data that is actually sent to the TBA port number? I think the text
> > > implies the
> > latter,
> > > which I assume is not correct.
> >
> > [Med] This applies to all messages that cross the converter.
> >
> > >
> > >
> > > Suggested text:-
> > >
> > > <<
> > >    This section defines the Convert protocol (Convert, for short)
> > messages
> > > that are exchanged between a Client and a Transport Converter.
> > >
> > >    Convert messages MUST be sent to TCP destination port TBA.
> > > Therefore,
> > a
> > > Transport Converter and a Client listen on this TCP port for Convert
> > > messages.
> >
> > [Med] The initial wording is correct.
> >
> > >    Convert messages MAY appear in a SYN, SYN+ACK, or ACK or MAY
> > > appear
> > in
> > > a subsequent packet.
> >
> > [Med] The initial wording is correct.
> >
> > > Convert messages MUST be included as the first bytes of the
> bytestream.
> > > All Convert messages start with a common 32 bits long header
> > > (Section 4.1), followed by one or more Convert TLVs (Type, Length,
> > > Value)
> > (Section
> > > 4.2).
> > > After a successful exchange of Convert messages, a TCP connection
> > > with
> > TCP
> > > extension(s) is established between the Client and Transport
> > > Converter (for instance, Multipath TCP), and a (normal) TCP
> > > connection is established between the Transport Converter and other
> > > end host, with the Transport Converter acting as an explicit proxy
> > > between the two connections (for instance, between MPTCP and TCP).
> >
> > [Med] No problem (even if the last sentence is already stated in
> > previous sections).
> >
> > > >>
> > >
> > >
> > > Section 4.0, 4.2.6 etc
> > > Various places say things like "the Unassigned field MUST be set to
> > > zero by the transmitter and
> > >    ignored by the receiver.  These bits are available for future use
> > >    [RFC8126]."
> > > Comment: I heard in ietf-105 about problems for extensibility of
> > > various protocols because implementations insist on all zeroes for
> > > fields, otherwise discard packets. The suggestion is to grease
> > > (which I think means that the senders set to random values and
> > > receivers MUST ignore)
> >
> > [Med] I don't see the value for doing this.
> >
> > > Also, 'sender' rather than 'transmitter'
> >
> > [Med] Fixed. Thanks.
> >
> > >
> > > Figure 11
> > > In the figure you have Value being optional in bits 16-31 and
> > > compulsory in bits 32+. I think this should be the other way round.
> >
> > [Med] OK.
> >
> > >
> > > Section 4.2.8
> > > "This TLV has a variable length.  It appears after the Convert fixed-
> > >    header in the bytestream returned by the Transport Converter."
> > > Figure 19 doesn't show variable length. Must its length be a
> > > multiple of
> > > 32 bytes (padded if needed)? (I assume so, to be consistent with
> > > elsewhere.)
> >
> > [Med] Agree. Fixed the figure.
> >
> > Padding is mentioned in the error description (when appropriate), e.g.,:
> >
> >      "The
> >       list of unsupported TCP options MUST be padded with zeros to end
> >       on a 32 bits boundary. "
> >
> > > The second sentence could be deleted, since elsewhere text says the
> > TLV(s)
> > > must be at the start of the bytestream. But if you keep the sentence
> > > I suggest you say "appears _immediately_ after"
> > >
> >
> > [Med] Deleted that sentence. No need to be redundant.
> >
> > > S6
> > > "The case of a middlebox that removes the payload of SYN+ACKs (but the
> > >        payload of SYN) can be detected by a Client."
> > > Do you mean: but _not_ the payload of SYN?
> >
> > [Med] Yes.
> >
> > >
> > > <<Appendix A.  Change Log
> > >    This section to be removed before publication.>> It would be
> > > really nice if somehow the material here that explains the design
> > > rationale, and development from earlier approaches, could be
> > kept.
> > > It's useful info, I think.
> >
> > [Med] OK, added a new appendix to cover the key points.
> >
> > >
> > > Best wishes,
> > > phil
> > >
> > > -----Original Message-----
> > > From: Scharf, Michael [mailto:[e-mail]]
> > > Sent: 22 July 2019 09:01
> > > To: Eardley,PL,Philip,TUD1 R <[e-mail]>
> > > Cc: [e-mail]
> > > Subject: WGLC comments addressed in draft-ietf-tcpm-converters-09?
> > >
> > > Hi Phil,
> > >
> > > Could you please have a look at -09 and let me know if your WGLC
> > comments
> > > are addressed?
> > >
> > > If not, please follow-up on the mailing list.
> > >
> > > Thanks
> > >
> > > Michael
> > >
> > > -----Original Message-----
> > > From: tcpm <[e-mail]> On Behalf Of
> > > [e-mail]
> > > Sent: Monday, July 22, 2019 8:04 AM
> > > To: [e-mail]
> > > Cc: [e-mail]
> > > Subject: [tcpm] I-D Action: draft-ietf-tcpm-converters-09.txt
> > >
> > >
> > > A New Internet-Draft is available from the on-line Internet-Drafts
> > > directories.
> > > This draft is a work item of the TCP Maintenance and Minor
> > > Extensions WG of the IETF.
> > >
> > >         Title           : 0-RTT TCP Convert Protocol
> > >         Authors         : Olivier Bonaventure
> > >                           Mohamed Boucadair
> > >                           Sri Gundavelli
> > >                           SungHoon Seo
> > >                           Benjamin Hesmans
> > > 	Filename        : draft-ietf-tcpm-converters-09.txt
> > > 	Pages           : 47
> > > 	Date            : 2019-07-21
> > >
> > > Abstract:
> > >    This document specifies an application proxy, called Transport
> > >    Converter, to assist the deployment of TCP extensions such as
> > >    Multipath TCP.  This proxy is designed to avoid inducing extra
> delay
> > >    when involved in a network-assisted connection (that is, 0-RTT).
> > >
> > >    This specification assumes an explicit model, where the proxy is
> > >    explicitly configured on hosts.
> > >
> > >
> > > The IETF datatracker status page for this draft is:
> > > https://datatracker.ietf.org/doc/draft-ietf-tcpm-converters/
> > >
> > > There are also htmlized versions available at:
> > > https://tools.ietf.org/html/draft-ietf-tcpm-converters-09
> > > https://datatracker.ietf.org/doc/html/draft-ietf-tcpm-converters-09
> > >
> > > A diff from the previous version is available at:
> > > https://www.ietf.org/rfcdiff?url2=draft-ietf-tcpm-converters-09
> > >
> > >
> > > Please note that it may take a couple of minutes from the time of
> > > submission until the htmlized version and diff are available at
> > > tools.ietf.org.
> > >
> > > Internet-Drafts are also available by anonymous FTP at:
> > > ftp://ftp.ietf.org/internet-drafts/
> > >
> > > _______________________________________________
> > > tcpm mailing list
> > > [e-mail]
> > > https://www.ietf.org/mailman/listinfo/tcpm
> > >
> > > _______________________________________________
> > > tcpm mailing list
> > > [e-mail]
> > > https://www.ietf.org/mailman/listinfo/tcpm
> >
> > _______________________________________________
> > tcpm mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/tcpm