A Concise Binary Object Representation (CBOR)-based Serialization Format for the Software Updates for Internet of Things (SUIT) Manifest
draft-ietf-suit-manifest-25

The shepherd writeup doesn't include the reason that PS is the right status here.

===

From Orie Steele, incoming ART Area Director:

In the abstract:

This would be clearer if the first use of "IoT" were
expanded.  (https://www.rfc-editor.org/materials/abbrev.expansion.txt
does not mark either as "well-known".)

Side note, is "Internet of Things Software Update (IoTSU)" a relevant / useful acronym here?

In Section 4.2 consider defining "pull parser", prior to using the term.

"The bootloader may add its own authentication, e.g. a Message Authentication Code (MAC), to the manifest in order to prevent further verifications."

Why?

In Section 5.3.4 "see {#ovr-integrated}," seems to be an escaped reference.

Section 5.4

Comment: "Severable Elements", seems similar to selective disclosure and elision, you might want to relate your definitions to other industry terminology.

Section 6.2

"""
If a Recipient supports groups of interdependent components (a Component Set), then it SHOULD verify that all Components in the Component Set are specified by one update,
"""

Under what circumstances can a recipient benefit from not verifying all components in the component set are specified by one update? (Why not MUST)

Noting Section 8.4.3 and Section 8.4.4 have internationalization considerations.

Section 8.4.8.10

'""
A URI Reference RFC3986 from which to fetch a resource, encoded as a text string.
"""

I believe this precludes URI's that contain unicode, you might consider borrowing some language from https://datatracker.ietf.org/doc/html/rfc9525#name-identifying-application-ser

"""
The information used to create the class-id condition (see {{uuid-identifiers)
"""

Reference typo.

Section "8.4.5.1. SUIT_Component_Identifier"

I'd prefer to see unicode considerations in this section.

in Section 11.8

Are expired Internet drafts considered acceptable "stable references" for "standards track range of point assignment"... If not, please give concrete guidance to the experts regarding managing registrations of references that expire.

"""
When specifications are not provided, the description provided needs to have sufficient information to identify what the point is being used for.
"""

Sounds like you are asking for registries that are "Expert Review" not "Specification Required", please confirm after reading:

https://datatracker.ietf.org/doc/html/rfc8126#section-4.6

Extra text in each registry section, would make all this a lot clearer, as I noted in the beginning of my comments.

In Section 11.9

Why not "application/suit-envelope+cose", under what circumstances can a suite envelope be transmitted as application/cose ?

If there will never be other expressions that are not COSE, "application/suit-envelope" seems fine.

Shouldn't [COSE_Alg] be a Normative reference?

Shouldn't all of the [I-D.ietf-suit-firmware-*] references be normative?
These are  optional, but if required/used, very normative on how to do it.
eg "if you want it encrypted, I-D.ietf-suit-firmware-encryption must be used".

        Finally, Appendix D gives a summarize of the
        mandatory-to-implement features of this specification.

That should not be in the appendix but in the main body of the document?

I find this text a bit confusing:

        [...] make the following guarantees:

        One of: 1. A [...]

        AND

        1.  Where

Might be the result of a misrender?

Section 6.4 uses a term "statement" not defined before. It really sounds
like "Command" to me, especially as one can "perform" a "statement".

Section 6.7:

        that Commands MAY be executed in parallel or out of order.

Should "or" be "and" ?

        Recipient may issue each or every command concurrently until
        the Strict Order parameter is returned to True or the Command
        Sequence ends.

This seems to invite race conditions if one concurrent command causes the
Strict Order parameter to change? What else could change the Strict Order
parameter midway a concurrent execution? Does this mean every command schedule
has to check the Strict Order parameter?

        When the manifest processor encounters any of the following
        scenarios the parallel processing MUST pause until all issued
        commands have completed

If the parallel processing pauses, wouldn't that pause the commands? I
think perhaps you mean "the parallel processing MUST not issue new commands
until all issued commands have completed" ?

        A Component MUST NOT be both a target of an operation and a
        source of data (for example, in Copy or Swap) in a Command
        Sequence where Strict Order is False.

Wouldn't this also be true when Strict Order is True?

Section 8.4.3

        suit-reference-uri is a text string that encodes a URI

What is a text string, esp. in relation to IDNs and non-ascii parts in a URI?

Section 8.4.11

Should suit-command-custom be called suit-command-custom-n with n a negative
integer? It might look from the command list there is only 1 custom command,
but there can be multiple ones.

Section 11.4 and 11.5 lists SUIT Commands/Parameters and shows a number
of Reserved fields in the middle of these new allocations. IS there
anything special with these values? Why are they reserved? Also, the text
states value 0 is skipped on purpose to detect "unset", so should value
0 not be listed as Reserved?

        A technique to efficiently compress firmware images may be standardized in the future.

Perhaps the compressing would be highly inefficient, to make decompressing highly efficient?
Maybe just remove the word "efficiently" :)

The CDDL contains:

suit-reporting-bits = &(
    suit-send-record-success : 0,
    suit-send-record-failure : 1,
    suit-send-sysinfo-success : 2,
    suit-send-sysinfo-failure : 3
)

Doesn't this mean 2 bits are used to declare a bool?
What if these contradict? Why not use a single bit?

NITS:

        see {#ovr-integrated},

Markdown misrender ?