i propose to enable/disable the option/features "ingesstpoint" and "streamalias" per acceptor/socket and not global.
Both feature will be very usefull in lot of cases for incoming and outgoing connections but generate a lot of extra work for features like recording and transcoding, as a streamalias make to be available before starting recording/transcoding. In case of "reboot" the streamalias is not existing anymore and recording (and transcoding) will not continue.
As well additonal datahandling is required to define a onetime streamalias to be able to record.
Additonal and switching of streamalias on/off will influence the commands required for recording&transcoding.
Perhaps a simpler implementation would be better? If we made ingest points optionally transient. Meaning that once an ingest point is used it will be deleted, much like aliases are now for outbound streams.
This way the stream is pushed in, assigned the localstreamname associated with the transient ingest point, and then the ingest point is removed. To push again to that same name, the ingest point would need to be recreated via API.
Would this satisfy your requirements?
it may satisfy some users, but not myself.
When using mosty rtmp ingest from different location on internet at same time there are a lot of reasons why a published rtmp get unpublished.
One of them could be internet cut, a reboot of encoder, …..
In this case the STream cannot be republished without trouble.
I believe there are some situation that a kind of "createingestalias" would be good. But I am searching for another solution which allows both of follwing needs.
a. a bit secured rtmp ingest from internet (external)
b. simplyfied complexity for backbone/internal usage (either on privat LAN or on CDN Backbone Infrastructure)
I’m thinking that we wouldn’t replace/remove the existing ingest points. They could be used for your internal/backbone network. You would then only provide public access to the "ingest aliases".
in my assumtion the "ingestalias" will be deleted when first "publishing" arrives. THis looks not good in my point of view. I see it as required to have a kind of "keepalive" for ingestalias. Either for a time (do not delete for *** seconds) or a kind of "keepalive" that a Alias is NOT destroyed automaticly.
In this case situation where RTMP get unpublished by fault for a short period will be no problem.
if client is not allowed to "push" anymore the "ingestalias" need to be deleted manual.
In current situation i don´t see a "ingestalias" comparable to a "addStreamAlias" as a helpfull solution for most cases. But maybe I missed some part of your idea.? Feel free to let me know.