[Emu] Re: WGLC for draft-ietf-emu-pqc-eapaka and draft-ietf-emu-hybrid-pqc-eapaka

[Heikki Vatiainen <[e-mail]>]

On Sat, 18 Jul 2026 at 18:26, Wang Guilin <Wang.Guilin=
[e-mail]> wrote:

- Section 7: A native attribute-level fragmentation mechanism is specified
> to address potential over-sized transportation of related attributes for
> enabling PQ KEMs, which is similar to the lock-step acknowledgment model
> used by EAP-TLS [RFC2716]. (I have a main comment below on this mechanism
> to check if a more efficient method is possible) .
>

Just noticed that the draft uses obsolete RFC number. RFC 5216, the RFC
that obsoletes RFC 2716, would be a better reference here.


> Main Comments:
>
> - On the attribute-level fragmentation mechanism defined in Section 7: It
> seems possible (and better?) to add one field or two fields for identifying
> the order of one Fragmentation attribute, like the 5th fragmentation out of
> total 10, in The Fragmentation attribute format specified in Section 7.1.
> This may become much more efficient to deal with one specific Fragmentation
> attribute lost. It may also allow for confirming multiple fragmentations.
> For example, one A sends all 10 fragmentations consecutively, B just
> replies with 9 to indicate the first 9 fragmentations have been received up
> to now, but the last fragmentation (#10) has not been received yet.
>
> The current lock-step acknowledgment model seems very strict and less
> efficient, I think.  It works like this: send one fragmentation => confirm
> receipt of it => send the next ...
>

I agree that this kind of batch sending of fragments may be more efficient.
Unfortunately I don't think is possible, though. The restriction comes from
the EAP specification which states that EAP is a 'lock step' protocol.

https://datatracker.ietf.org/doc/html/rfc3748#section-2

With regard to draft-ietf-emu-pqc-eapaka section '7.3. Use of the EAP
Identifier', I would remove the section from the draft. The EAP layer
already does what the section defines. Leaving it in would be confusing.
EAP's 'lock step' behaviour does everything that's needed and EAP-AKA'
needs just to send a receive its messages with fragments and fragment
acknowledgements.

-- 
Heikki Vatiainen
[e-mail]