[linux-audio-dev] What parts of Linux audio simply suck ?

Juan Linietsky coding at reduz.com.ar
Mon Jun 20 00:43:36 EDT 2005


ix at replic.net wrote:

>>As I said many times It's not that I hate Linux Audio, but mainly that I 
>>    
>>
>
>mmh, i love it,
>
>  
>
>>believe that it is too poor
>>in features/API for my taste. I tried very hard for several years to 
>>make my apps play together
>>with jack/alsa, but I find myself very limited in many areas as a 
>>programmer and user:
>>
>>-Alsa/Jack integration in timestamping is poor, syncronizing audio to 
>>midi is a pain
>>    
>>
>
>so use Jack MIDI
>
>  
>

Just found it it's been released! Awesome!

>>-I have no way to ask from a sequencer, the patch names available in a 
>>softsynth to list them
>>    
>>
>
>i do this all the time, grabbing lists of automatable params and patches from my sequencer, using OSC aware apps like PD, Om etc as the LADSPA/DSSI hosts. sounds like a sequencer limitation if it's not exposing these things programatticaly for you..
>
>  
>
OSC aware apps? wait, wait wait..  you can for sure use osc, but using 
jack and osc separatedly?
That is like, a bad hack.. I mean, how do you know which host is what? 
Much less sync properly
(read answers below)

>>-Jack lack of midi
>>    
>>
>
>false..
>
>  
>

Right, Just checked on jack ML.. this is pretty amazing!

>>-Jack lack of OSC or any way to do parameter automation from the sequencer
>>    
>>
>
>im glad jack concentrates on audio and not "OSC parameter automation". its the UNIX way, a utility doing one thing and doing it well, but if Jack could deliver sample-accurate OSC messages in a more convenient manner than UDP, im all for it, but the UI thing you speak of is a level above that.
>  
>
What are you talking about? without frame-based precision you are 
relying on wathever jack server
period size is, which means unreasonably forcing your apps to _really_ 
low latency for just "decent" accuracy.
It is the same problem you have when you use alsaseq+jack. The typical 
hack to fix this is to have a separate thread receiving and timestamping 
events, but this is plain SHIT because you have to anyway rely on the 
lowlatencyness of the kernel to not jitter the alsa/osc IPC.
I'm _NOT_ glad that jack concentrates only on audio, because I cant do 
that, share tempo maps, and many of the things described here.

>  
>
>>-It is Impossible to do any sort of offline render, or high quality 
>>render of a song (like, in 32/192khz) using JACK/Alsa
>>    
>>
>
>so you want to be able to run the JACK samplerate at 'infinite' or at least a high as possible until processing finishes. an interesting idea, to have offline render across the entire chain.. but last i asked about even so much as multiple samplerates or vari-speed scrub across all jack spps,  they had reasons, not the least of which was it would be useless without everyone supporting it..
>  
>

Come on!! *ANY* Comercial grade sequencer/daw/etc supports offline 
render. Like for example when you want to compose for movie/DVD, you 
need to use very high sampling rates/channels/etc that if you try to 
play in realtime would crawl your computer to death. So you just offline 
render the whole thing. But for more normal-composer uses, you probably 
want to render in high sampling rates and then downsample to reduce 
aliasing produced by several dsp techniques. Currently this is 
impossible with jack.

>  
>
>>-Saving/Restoring your project is just painfully hard. LASH doesnt help, 
>>and even when I came up with the idea of it in the first place.
>>    
>>
>
>this could be improved, but monolithic Cubase-esque hosts is not the way id prefer it to be done. personally i just have my sequencer spawn off a few shell processes coupled with rohan drape's excellent patching utils to handle app launches, while storing all the params&patches id want to recall in the sequence database..
>
>  
>
I dont really mind if it is monolithic or not, I'm just noting that I 
need it, and I cant do it.

>>-Adding/Removing softsynths, linking connections, etc takes a while 
>>having to use qjackctl, etc
>>    
>>
>
>''
>  
>
Sure, you have to go and connect everything yourself, every time you 
start up, go to the shell
find an empty console, run fluidsynth or wathever, have everything 
organized, etc. Basically
the same as in the above point.

>>-Lack of send%.. I just cant have a jack client doing a very high 
>>quality reverb, only as wet processing and have clients send different 
>>amounts of the signal to it, thus saving CPU
>>    
>>
>
>so you want jack to do mixing + connections, instead of just connections. take it up with them, but variable wet/dry requires a 100% connection, so im not sure thats not a UI issue a level above jack too..
>
>
>  
>
Well, seeing how jack works i'm not sure if it does zerocopy because at 
some point it has to add up the connectons together, so I dont think 
adding sends per connection is impossible. Yes, ican probably make
an inbetween proces that does a send, but this is annoying to use and 
probably reduces performance a lot more too.

>>-Lack of tempo-map based transport, I cant adapt my midi-only sequencer 
>>, which works in bars,
>>beats, etc to a frame-based transport.  Say I want to run my sequencer, 
>>then go thru softsynths
>>and record/edit in ardour.. no go.
>>    
>>
>
>confused here.. as to why it wouldnt work, other than your unwillingless to make a simple function to cnovert between beats/bars to samples..
>
>  
>
Because ardour for example, also has it's own tempomap and what I record 
on it wont fit. If i then
change the tempo of my song or decide to do other changes 
(signature/etc), everything screws up.

>>But overall, what mostly annoys me of linux audio is that most API 
>>programmers just implement
>>the features THEY use and need, and not what others may need. And since 
>>they mantain the thing,
>>    
>>
>
>is this not what the economics dictate? if someone sat down with major backing to implement a 'grand vision' API, wouldnt we be talking about Apple?
>
>  
>
>>even adding them yourself is pointless, as they will most certainly not 
>>accept patches. Ok, that's
>>    
>>
>
>not true, even my small trivial bugfix and usability patches have been incorporated into various projects.. if youre talking about something enormous, maybe a branch is better at least until its utility/superiority/whatever is proven..
>
>  
>
>>fine, they are on their right to do it, after all i'm not paying them to 
>>do it, they work for themselves.
>>
>>All this has simply led me to decide to not use such APIs anymore and 
>>integrate everything I do
>>in big, and monolithic apps, such as reason, cubase, etc and not care 
>>    
>>
>
>have fun, im personally glad to get away from that..
>
>  
>
Yes, see what I mean, you have your point of view on the subject, so do 
API/lib/app developers.
You could just go and argue about a lot of topics like:
-Should jack support OSC?
-Should apps be monolithic or a modular set of them?
-Should audio/midi plugins separate UI from core in two processes or 
should it be integrated?
-Should transport work frame based or BarBeatTick/etc?
-Should we make a more advanced plugin standard and adapt it to an app, 
or wait for GMPI?
-Should we use plugins at all?
-Should we advocate for low latency, or better timing/syncronization so 
low latency is not needed as much?
-etc, etc

Everyone has their point of view. It's not like you will tell someone "I 
want to add this feature to your app/api" and will say "Ok". You will 
simply get an answers like:
-No, sorry, I wont accept that patch, i'd rather the library 
concentrates only on this.
-Why dont you do it as a separate library?
-Feel free to fork this and add it yourself.
-Yeah I recognize it's useful, but I think it's out of place, 
inconsistent with the rest, that I try to keep simple.

etc, etc, etc.

And forking of course is not an option. I mean it's a lot of work 
because you have to fork the lib, mantain it, recompile all the apps 
that use it against yours and not the main one, merge bugfixes and 
improvements to keep compatible. Also if you are using a distro, you 
cant use the binary packages because they are not compiled with your 
version of the lib, etc. So forking is not really worth it in most cases.

>>about the outside world anymore.
>>After all, it takes me less time to write the features I need for 
>>myself, and into my own than dealing with people's religious software 
>>views to get them integrated into other projects.
>>    
>>
>
>wish i could code that fast!
>
>  
>

Well, I dont have much time, nor I code very fast, but it's more like.
I made Sequencer A, and softsynth B. They ar communicated thru library
X. I want Sequencer A to do the things I want with softsynth B, both are 
my work.
But library X doesnt support, it, so I go talk with library X developer 
and it seems
there are no chances of that being implemented in the near future, or 
ever. So
what do I do? I'm not left with much choice, Either I fork library X, or 
make
A and B a monolithical app. The second is the easier/faster so I go for 
that.

Juan




More information about the linux-audio-dev mailing list