SRT

[funman]

SRT protocol

[funman]

I hacked rist_ts and rist_rx changes to test quickly

Not sure if I should remove the changes completely or add new programs

[funman]

This depends on https://github.com/funman/bitstream/tree/srt

[nto]

I hacked rist_ts and rist_rx changes to test quickly

Not sure if I should remove the changes completely or add new programs

You should probably add srt_tx and srt_rx.

[funman]

@funman srth: no need to update flow def immediately breaks data transmission

This seems to break rist_tx caller

[nto]

There are still some indentation changes left in rist_rx and rist_tx.

[funman]

There are still some indentation changes left in rist_rx and rist_tx.

Sorry I think I used git diff -w

[nto]

What about adding upipe_srt_source and upipe_srt_sink pipes which could be used as replacement of upipe_udp_source/upipe_rtp_source and upipe_udp_sink? These would make srt_tx/srt_rx much simpler, and srt usage in client applications as well.

[funman]

What about adding upipe_srt_source and upipe_srt_sink pipes which could be used as replacement of upipe_udp_source/upipe_rtp_source and upipe_udp_sink? These would make srt_tx/srt_rx much simpler, and srt usage in client applications as well.

I don't understand what they would replace exactly, why does udpsink need to be replaced?

[nto]

I don't understand what they would replace exactly, why does udpsink need to be replaced?

The same way you could replace upipe_udp_source with upipe_rtp_source to use RTP instead of UDP, you should be able to use upipe_srt_source (with parameters passed in the URI) to use SRT instead.

[funman]

I don't understand what they would replace exactly, why does udpsink need to be replaced?

The same way you could replace upipe_udp_source with upipe_rtp_source to use RTP instead of UDP, you should be able to use upipe_srt_source (with parameters passed in the URI) to use SRT instead.

I guess it would have to handle the creation of udpsink and srt handshake as well then?

[nto]

Yes, it encapsulates everything, which is why srt_tx and srt_rx will be much simpler afterwards.

[funman]

It's not simpler for me since they are done already :)

[funman]

Yes, it encapsulates everything, which is why srt_tx and srt_rx will be much simpler afterwards.

Also wouldn't we be missing events between internal pipes that could be useful?

[funman]

@kierank any insight here?

[kierank]

Maybe once we have fixed the bugs we could have a wrapper pipe?

[nto]

Maybe once we have fixed the bugs we could have a wrapper pipe?

Agreed. It does not have to be in this pull request.

[kierank]

@cmassiot @nto @quarium I think this is "done"

[funman]

No, I think it should be on its own

Le lun. 9 sept. 2024, 13:13, JDarnley @.***> a écrit :

@.**** commented on this pull request.

In examples/rist_rx.c https://github.com/Upipe/upipe/pull/947#discussion_r1750062444:

@@ -331,7 +331,7 @@ int main(int argc, char *argv[])

 udp_sink_mgr = upipe_udpsink_mgr_alloc();

• uclock = uclock_std_alloc(UCLOCK_FLAG_REALTIME);

• uclock = uclock_std_alloc(0);

Did you intend this change and the similar one in rist_tx to remain in the PR?

— Reply to this email directly, view it on GitHub https://github.com/Upipe/upipe/pull/947#pullrequestreview-2289599154, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAVVXK2KSORXARENRULQDLZVV7FXAVCNFSM6AAAAAA5V4S66GVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDEOBZGU4TSMJVGQ . You are receiving this because you were mentioned.Message ID: @.***>

[kylophone]

What is the status of this SRT implementation? Intended to be mainlined at some point? Experimental?

[kierank]

Hi Kyle,

I have been using this implementation in production for nearly a year now without major issues and interoperating with various libsrt library versions.

Most of the recent work has been around a few edge cases for rogue callers. The intention is to mainline it after some review. It's worth saying it's hard to unit test and also fuzz a complex protocol like SRT so that's a missing piece of work.

Regards, Kieran Kunhya

On Wed, 22 Jan 2025, 21:42 Kyle Swanson, @.***> wrote:

What is the status of this SRT implementation? Intended to be mainlined at some point? Experimental?

— Reply to this email directly, view it on GitHub https://github.com/Upipe/upipe/pull/947#issuecomment-2608329871, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABDEEGZWL4UELAIPJYU5KD2MAGFHAVCNFSM6AAAAAA5V4S66GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMBYGMZDSOBXGE . You are receiving this because you were mentioned.Message ID: @.***>

[kylophone]

Sounds good. Thanks for the info, Kieran.