This document has been reviewed as part of the transport area review team's
ongoing effort to review key IETF documents. These comments were written
primarily for the transport area directors, but are copied to the document's
authors and WG to allow them to address any issues raised and also to the IETF
discussion list for information.

When done at the time of IETF Last Call, the authors should consider this
review as part of the last-call comments they receive. Please always CC
tsv-art@ietf.org if you reply to or forward this review.

It was very difficult to review this document for IETF transport protocol considerations. 

Although "transport encapsulation" is indicated repeatedly, it is never referred to directly or described either in this document or its citations. It appears to be using this term in the sense of RFC8300, which too never defines it, but uses examples that are more accurately referred to in the IETF as link layer protocols or either network or link tunnel protocols (IP in IP, GRE, VXLAN, Ethernet).

Regardless of the fact that this confusion originates in RFC8300, it needs to be addressed here and corrected before this document can be reviewed to determine if there are any IETF transport area issues.

The remainder of these notes provide detail of this issue.

-----

The document refers back to RFC8300 to define the NSH itself; that document discusses transport issues just as vaguely (never mentioning a particular transport protocol), and when it discusses fragmentation, it refers to section 9 of a document (draft-ietf-rtgwg-dt-encap-02 from 2017) that had expired prior to the publication of RFC8300.  Because transport fragmentation is, IMO, a normative issue, this should not have been permitted.

Further, Section 9 of that draft incorrectly recommends reliance on ICMP feedback to address MTU failures when not under a single operator’s management. That was widely known even then to be insufficient due to blackholing; this had motivated PLPMTUD in RFC4821 a full decade earlier. RFC8300 compounds this error by simply asserting that the operator should ensure that ICMPs are not blocked, overlooking the need to address when this is not the case.

This document cannot ignore that issue and simply refer to RFC8300 on this issue.

Note that one of the only places an actual encapsulation protocol is mentioned is RFC8300, in which Section 5 mentions IP and  Section 6.1 Table 1 describes VXLAN-GPE, GRE, and Ethernet – all of which are described as “transport encapsulation”. 

If, in fact, IETF transport protocols are being used, at some point the use of an actual IETF transport protocol should be described (e.g., TCP, UDP, SCTP, DCCP). At that point, the transport issues would be reviewable. As the document currently stands, it completely ignores such transport issues and should not proceed until this is addressed and re-reviewed.

If instead, as I suspect, the term “transport encapsulation” actually refers to “network layer encapsulation” or “link layer encapsulation” and really implies some sort of tunnel, there would be no transport area issues to review unless that tunnel were to include a transport protocol as part of the layers of encapsulation. If that is the case, the document should be revised to replace the term “transport” with something that more accurately describes VXLAN-GPE, GRE, Ethernet, and IP encapsulation using IETF terminology. Note that draft-ietf-intarea-tunnels never uses the term “transport” except when referring to the use of IETF transport protocols as a tunnel layer, e.g. (i.e., the last sentence of Sec 8 of this doc is incorrect in implying otherwise).

(I would also note that neither this doc nor RFC8300 define “transport encapsulation” in their terminology; even if they would, they should not attempt to define it in a way inconsistent with widespread use in the IETF).