phil-news-nospam@ipal.net wrote:
> In comp.protocols.tcp-ip "Nils O. Sel?sdal" <NOS@utel.no> wrote:
> | phil-news-nospam@ipal.net wrote:
> |> I've read about and downloaded the packages sctplib and socketapi which
> |> appears to use sctplib. But these are userland implementations of SCTP.
> |>
> |> What I'm really looking for is how a non-userland implementation would
> |> be used by a userland application program. I presume this would use the
> |> basic socket interface. But there are details about SCTP that would also
> |> make this differ somewhat. For example, identifying different streams.
> |>
> |> What is the expected standard way (API) to use SCTP for non-userland (e.g.
> |> in the host kernel IP stack) implementations of SCTP?
> | http://lksctp.sourceforge.net/ is a nice starting point
> | Examples, standard API drafts and others are linked from there.
> |
> | sctplib has an addon giving you the socket api(well, almost),
> | described in
> | http://www.ietf.org/internet-drafts/...psocket-11.txt
>
> Is that the API that will be used for kernel socket syscall based SCTP?
Yes.
For linux, some of the extended calls are placed in a library, again,
see http://lksctp.sourceforge.net/

> I'm going to be enabling SCTP in my Linux kernel to see how well that
> works. I suspect a purely library in userland based method will incur
> access restrictions (I presume it needs raw sockets) and that can be a
> problem with user applications that will use SCTP.
Yes.

> Anyone have experience programming around either method (preferably for
> other than SS7 or telephony signaling purposes)?
That question is too vague..