Re: [Anima-bootstrap] bootstrap over CoAP

[Brian E Carpenter <[e-mail]>]

On 10/07/2016 01:46, Carsten Bormann wrote:
>> 1. Is CoAP/DTLS protected against corrupt packets? (Assuming we are
>> talking about UDP/IPv6 there will at least be the UDP checksum.)
> 
> DTLS ciphersuites usually authenticate the packets (and protect against
> replay), so there will be very strong protection.

Good, that's what I hoped to hear.

> 
>> 2. In the fragmentation scenario, what happens when a fragment is
>> corrupted or lost?
> 
> I'm not sure I understood that part of the draft*), but generally CoAP
> is designed so you can avoid fragmentation 

That isn't what the draft seems to say, though. However, it's clearly talking
about application-layer fragmentation to avoid IP fragmentation.

> (and use the segmentation
> provided by draft-ietf-core-block instead); the latter has per-segment
> reliability (acknowledgements and retransmits).   DTLS may require
> fragmentation during its handshake; this is mitigated if you can use a
> PSK (symmetric) or, if you need asymmetric, ECC-based ciphersuite, which
> allows the packets to stay well below 1280 bytes.

Right. So the app layer chops the message into <1280 byte pieces, and one
of them is lost...?

Regards
    Brian

> 
> Grüße, Carsten
> 
> *) We just had an interesting exchange in the CoRE WG where some people
> weren't aware of the terminology that differentiates fragmentation and
> segmentation; I apologize for sticking to that.
>