On Tue, 26 Nov 2013, Remy Leone wrote:

> If the feedbacks and interests are present, we might dedicate an internship
> for making it possible. Right now we are probing for interest, directions
> and requierements.

From what I can tell, it seems to fit the pattern: an application layer
internet protocol for transfers, whose "targets" can be specified with a
URL/URI.

> - Is it mandatory for a protocol to have the RFC status in order to be
> integrated into libcurl?

No, but it helps. We're an informal bunch of people, we have virtually no
strict rules that can't be reconsidered or negotiated given a good motivation.

For non-RFC protocols I would say that it helps if the protocol is popular or
widely used. This doesn't seem to fit that description - although it might
also just be an area I've not explored much personally.

> - How many unit tests are required for a new protocol to be integrated?

There's no limit, but the more the better. What tends to be the most hard part
when adding tests for new protocols is to write/provide a test server with
enough flexibility so that we can make the test suite control it properly. A
test server needs to be able to also violate the protocol suitably when asked
to so that negative tests can be made.

Unfortunately we don't even have test servers for all protocols we support.

> - Do you have walkthroughs or tutorials in order to implement a protocol
> into the libcurl?

No we don't. Possibly you can help us write one based on your experiences here
with this!

But I would recommend that you for example check out the TFTP protocol handler
since that is also UDP. Or perhaps scp/sftp that are very state-machine
intensive and use a 3rd party lib for the binary code/decode parts of what
flies over the connection.

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html