Hello Peter,

On Tue, Feb 15, 2011 at 12:32 PM, Peter Walter <[email protected]> wrote:
> as you said that splitting liquidsoap into multiple instances might be a
> good idea anyway, there's one more thing regarding marshals. first you say
> that the technique had disappeared from SVN (and it's no more mentioned in
> the beta or SVN docs), but in the last paragraph you tell me that i could
> use fixes on marshal with an SVN upgrade? it was amazingly simple to connect
> liquidsoap processes that way, nothing in between but a named pipe. if a
> solution is or will be available where a broken pipe doesn't kill a process,
> then i would be very interested.

To clarify, I meant that if marshall I/O seems necessary we can
restore it. I won't make the effort if there are safer solutions
(marshalling has inherent limitations) and I wouldn't want anybody
doing it if it's not cleaned up at the same time (it should notably
support failures). Now, I'm not sure how necessary it is because we
now have raw streaming from one liquidsoap to the other using
output.http and input.harbor. Compared to harbor, metadata may be lost
but there are easy ways to fix this. There might be other differences,
but I'd need to look into the technical details to know what was
really supported. In any case, it'll probably never be as good as a
source-to-source connection within a single liquidsoap, but it
shouldn't matter for most people.

Also, having multiple liquidsoap processes is indeed interesting, but
it should only be needed to deal with crashes (which don't happen, at
least in theory) and restarts. Latency issues should all be taken care
using clocks, at least that's the hope.

Looking forward to hear from your experiment with clocks,
-- 
David

------------------------------------------------------------------------------
The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE:
Pinpoint memory and threading errors before they happen.
Find and fix more than 250 security defects in the development cycle.
Locate bottlenecks in serial and parallel code that limit performance.
http://p.sf.net/sfu/intel-dev2devfeb
_______________________________________________
Savonet-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/savonet-users

Reply via email to