NFD » ndn-cxx

Reminder sent to assignee: https://www.lists.cs.ucla.edu/pipermail/nfd-dev/2019-October/003828.html

For the deriveCertBundleName and deriveCertBundlePrefixFromKeyLocator functions, what should they return if the name passed to the function isn't correct?

what should they return if the name passed to the function isn't correct?

If argument is wrong, throw std::invalid_argument exception.

Part of this task is blocked by #4696 because encodeCertBundle relies on SegmentPublisher.
Other parts are not blocked. You can submit a Gerrit Change when unblocked parts are ready.

Reminder sent to assignee: https://www.lists.cs.ucla.edu/pipermail/nfd-dev/2019-November/003874.html

Should I ignore encodeCertBundle for now, write the skeleton of the method around the call to the SegmentPublisher, or include a simple segmentation implementation within it?

Should I ignore encodeCertBundle for now, write the skeleton of the method around the call to the SegmentPublisher, or include a simple segmentation implementation within it?

You should submit a Change without encodeCertBundle and anything else that depend on SegmentPublisher.
You should also push the assignee of #4696 to complete his/her assignment, so that you can continue working on encodeCertBundle.
In either case, "simple segmentation implementation" is NOT part of this task.

Since the bundle decoder is a stream decoder, when a data segment is appended, should the decoder check if that segment is (or is part of) a certificate?

Does it make each bundle available as they are decoded?

when a data segment is appended, should the decoder check if that segment is (or is part of) a certificate?
Does it make each bundle available as they are decoded?

Yes. The logic would be similar as reading top-level TLV packets from a stream socket.

Is it possible for a data segment to contain part of two different certificates? For example, if segments are 8000 bytes each and a certificate is 9000, would the first certificate be contained in the first and second segment and the start of the second certificate also be contained in the second segment?

If this is the case, is there a class in ndn-cxx for reading TLV data directly?

Is it possible for a data segment to contain part of two different certificates? For example, if segments are 8000 bytes each and a certificate is 9000, would the first certificate be contained in the first and second segment and the start of the second certificate also be contained in the second segment?

Yes, this is allowed by protocol.

is there a class in ndn-cxx for reading TLV data directly?

SegmentFetcher and SegmentPublisher are expected to expose a streaming interface. Decoding and encoding certificates in the payload stream of a segmented object shall be similar as decoding and encoding L3 packets in a Unix socket.

is there a class in ndn-cxx for reading TLV data directly?

SegmentFetcher and SegmentPublisher are expected to expose a streaming interface. Decoding and encoding certificates in the payload stream of a segmented object shall be similar as decoding and encoding L3 packets in a Unix socket.

Could you please point to the code?

I don’t know enough about C++ version. In Python it’s here:
https://github.com/zjkmxy/python-ndn/blob/16cb6de6537a334ada2e76c2a43b156269edfed3/src/ndn/transport/stream_socket.py

What is the SegmentFetcher exposing its streaming interface to? Is the purpose of the append function in the CertBundleDecoder not to accept incoming segments from the Fetcher?

What is the SegmentFetcher exposing its streaming interface to? Is the purpose of the append function in the CertBundleDecoder not to accept incoming segments from the Fetcher?

CertBundleFetcher would use SegmentFetcher to retrieve the segments, and pass each received segment to CertBundleDecoder::append.

Ok. What do you mean by "expose a streaming interface?"

Ok. What do you mean by "expose a streaming interface?"

This means the decoder can incrementally accept input, and emit successfully decoded certificates as they are received.
In TypeScript analogy, it's a function that accepts AsyncIterable<Uint8Array> type and returns AsyncIterable<Certificate> type.

The design states that "A producer MAY provide one or more certificate bundle names in the RDR metadata."

Should creation of the metadata packets be part of the publishing API or is my goal here just to extend the existing MetadataObject with the ability to create such packets?

MetadataObject needs to have an extension for Certificate Bundle name, but publishing RDR metadata isn't certificate bundle's responsibility.