Hi,

RingTeam wants to provide file transfer. And we want to do it fast! How fast? 
Saying in the order of a month.
I've opened a Redmine ticket to track that: 
https://projects.savoirfairelinux.com/issues/80747

I'm going to report my design progress using this ML and I open the discussion.

For my first report I see these facts:

- the solution has to work on our current provided accounts : IP2IP / RING / SIP
- the solution needs to provide NAT-traversal
=> we already have a technology for that: ICE. But our implementation requires 
to use UDP transport
- the solution has to provide handling API to request/start/stop file transfer
- the solution has to provide meta-data display starting at request, to let 
recipient decides what to do (accept/refuse)
- the solution has to provide authentification/encryption 

I've already identified some related protocols to help to solve this task (see 
ticket associated tasks): UDT over SIP/SDP, rfc5547/rfc4975, tftp, QUIC?

As I want to keep the development in a relative short time and the an easy 
integration, I think that rfc5547/rfc4975 and QUIC are not the right solutions 
(too heavy to integrate).
tftp is too simple and restritive for our needs.
UDT is right for data transport but doesn't bring 
meta-data/session/authentification/encryption (need to be coupled to SIP/SDP 
and nettle/gnutls for authentification/encryption)


This last is promising, re-using most of our already integrated technologies, 
but some refactoring remain if we want to keep transfer outside a call context 
(where the SIP/SDP session live).

Any comments/ideas?

--
Guillaume R.
Ring Project Director
_______________________________________________
Ring mailing list
[email protected]
https://lists.savoirfairelinux.net/mailman/listinfo/ring

Reply via email to