Re: What's the issue with SCTP-based data channels on servers?

Do you mean SCTP or SCTP over UDP ?


> On Oct 23, 2018, at 11:05 AM, Lennart Grahl <lennart.grahl@gmail.com> wrote:
> 
> Hi everyone,
> 
> we are repeatedly being told that *deploying* SCTP-based data channels
> is too cumbersome. But I have yet to find out what's so cumbersome about it.
> 
> There are libraries that terminate SCTP-based data channels (in fact,
> I'm the maintainer of one particular implementation written in C called
> RAWRTC). Even though I wouldn't exactly call my implementation very
> mature, so far I have not heard of any deployment issues. The only thing
> that could be slightly related is an issue with cross-compiling... but
> that's pretty much it. Though, might be a popularity problem.
> 
> I have overheard statements that usrsctp is immature. Please, if you
> believe so, explain why. I give you that much: It uses an awkward API
> since it moves the socket API into userspace. But that doesn't make it
> immature. Furthermore, you don't need to use its API directly since data
> channel implementations abstract it away.
> I've also heard that usrsctp is hard to deploy: If so, I would like to
> know that the issue is.
> 
> There have been statements that there aren't any modern SCTP
> implementations, that it is an old protocol and there generally is no
> interest in it. That is not true, in fact there are people who have
> written one just for the purpose of using it in WebRTC: aiortc (Python
> 3), pions/webrtc (Go). I know of at least another person writing one in
> Rust and Tim has written one in Java.
> 
> So, to come back to my original question: What are the actual
> *deployment* problems people face? If there is no deployment issue, what
> is it that drives you away from SCTP-based data channels?
> 
> If you do have contacts that might be able to provide feedback, please
> forward them this mail.
> 
> Cheers
> Lennart
>

Received on Wednesday, 31 October 2018 22:11:24 UTC