The Author of the paper is advocating using pulse/jack rather than using Alsa directly i.e. using a client server model rather than a sound device is a file model. For jack in particular - having a pull based model work efficiently is very important. The Author notes that writing apps the OSS way encourages apps that don't necessarily play well with the rest of the sound system. You can certainly make a case that cross platform is important but you cannot simply ignore these other factors. For jack you get significant benefits for your application using jack over using alsa,oss directly - I imagine it is the same for using pulseAudio."Stop developers from writing apps with the OSS API"
Yep, true cross-platform friendliness. Write for ALSA so that your app will work on Linux, BSDs, etc. Oh no, ALSA is Linux-only. First you blame MS for such tactics, but then you go and embrace them? If you write for OSS, your app will work on many Unices, not just Linux. So why use ALSA in the first place?
The paper was written in 2009, When was the modern version of OSS made publicly available?Oh, and didn't anyone notice that the PDF comes up with the shortcomings of OSS that only exist in the ancient in-kernel version and don't exist in the modern implementation? How convenient, don't you think?
It should give some insight as to where Linux audio is going and why.
At this point it is most important to have a model that works and has the manpower behind it to fix the things that don't work. Pulse is a reasonable canditate for this. OSS could have been a candidate but because of the bridges that the OSS team burn't in the past I really can't see everybody moving that way any time soon. (If OSSv4 had been freely available before everybody migrated to pulse-audio there is a very small chance that things might have been different)