Re: [dtn] DTN BPv7 test bundles and CRCs

Hi Brian,

This sounds really useful! I agree that it would be very good to have
some example bundles provided. We have Python as well as C code for our
interpretation of the BPbis CRC section available with µPCN v0.6.0 [1].
The Python implementation can be found in tools/bundle7/crc.py, the C
implementation in components/bundle7/src/crc.c.

Please note that in this version the used CRC-32 variant is still the
PKZIP/Ethernet CRC-32 as specified in draft-ietf-dtn-bpbis-10. For using
CRC32-C (Castagnoli), the table in the C file needs to be re-generated
with the corresponding polynomial. This can be done using the Python
code which includes parameters for a lot of CRC variants.

Felix

[1] https://upcn.eu/

--
Dipl.-Ing. Felix Walter
Chief Technology Officer
D3TN GmbH

E-Mail: felix.walter@d3tn.com
Web: https://d3tn.com

Address:
D3TN GmbH
Leon-Pohle-Str. 2
01219 Dresden

Commercial Register: District Court Dresden
Register Number: HRB 34957
VAT ID No.: DE303348407
Managing Directors: Marius Feldmann, Paul Seidler

On 22.03.19 16:45, Brian Sipos wrote:
> To anyone who has been prototyping Bundle Protocol version 7 (BPbis)
> agents,
> I'm spending a little time on a Wireshark dissector to examine and
> troubleshoot bundles and the only trouble I've run into is in
> verifying the CRCs. The BPbis spec is not very helpful in determining
> exactly what CRC implementation to use only because the ITU
> specification itself doesn't provide CRC test vectors.
>
> I'm doing prototyping in Python and dissecting in C language, so there
> is even a disparity between what CRC algorithms are available and how
> well they are documented. This seems like such an easy use/build
> decision to use COTS CRC function but I'm struggling to find
> implementations to match what BPbis requires.
>
> I think there would be a lot of value to have some test/representative
> bundles in an appendix which include blocks with different CRC types
> and different block sizes. Some CRC implementations pad out
> octet-sequences to a multiple of the CRC size and some apparently do
> something else. The problem is in the CRC implementations not being
> very clear about their parameters but having a few encoded test
> bundles to work with would go a long way to ensuring compatibility of
> implementations.
>
> Thanks for any advice,
> Brian S.
>
> _______________________________________________
> dtn mailing list
> dtn@ietf.org
> https://www.ietf.org/mailman/listinfo/dtn