Data segmentation

[aelhadee]

Hello,
I have a quick question, has the data segmentation for data larger than 8800 bytes been implemented in this repo or it needs to be incorporated manually?

Thanks,
AE

[zjkmxy]

Python-ndn does not do segmentation autmatically.
Actually, python-ndn is not aware of this limit, because this is required only by NFD. NFD will drop your Data packet if it is larger than 8800B. This is applied to Data packets after fragment reassembly, so the library generally cannot help you do anything.
I recommend you to segment your data payload in a size of about 4~6kB.

[aelhadee]

Hello @zjkmxy, I have tried data segmentation with python-ndn at both the sender and receiver and was not successful. Is there a sample segmentation code that I can refer to at both devices?

[zjkmxy]

For the producer, you can refer to putchunks.py.
For the producer, you can refer to catchunks.py.

You can also refer to rdrnode.py, which implements the RDR protocol and acts as both producer and consumer. However, this example uses NTSchema API, which is experimental and may change frequently.

[aelhadee]

These really helped! Thank you!

For the producer, you can refer to putchunks.py.
For the producer, you can refer to catchunks.py.

[aelhadee]

Hi @zjkmxy, I have tried to use putchunks to send fresh data by using data = os.urandom(32000) so it would send new data every time there is a new interest from catchunks, but it didn't seem to work. It seems like the same data is sent over cnt times. Any idea why that is happening?

[zjkmxy]

In NDN different data are supposed to have different names. If catchunks sends the same Interest again, it will hit NFD's cache instead of the producer.

[aelhadee]

So for some reason when the time stamp is added to the name in putchunks.py, the catchunks.py does not detect it and the interest times out. Is there a reason for that? It only works when I remove the timestamp and just use the name by itself in putchunks.py.