One other note: in Jack1, to aid with "back compatibility", if you were to use -Xseq, Jack1 will interpret that as "use my embedded version of a2jmidid".

as noted by others, unless you do not care about timing, you should never use -Xraw. You should use -Xseq only with Jack1, where it is equivalent to starting up a2jmidid -e, which is the recommended approach for Jack2.

On Mon, Jun 18, 2018 at 9:55 AM, Victor A. Stoichita <victor@svictor.net> wrote:

Le 18 Jun 2018, Paul Davis <paul@linuxaudiosystems.com> a écrit :
Just a note. You should NEVER use ALSA raw MIDI with JACK unless you don't care about timing in any way.
(…)
And for those wierdos/smart folks/rebels out there using Jack1, remember that a2jmidid is builtin to Jack1 itself.

Thanks Paul for the hints! I do care about timing, that’s why I’m researching this. I think that I may run into some jitter when recording midi in Ardour or Qtractor (or I really play worse than I thought!). That’s what got me to the advice of using -Xalsarawmidi combined with a2jmidid. I’m still not sure whether this is the same as -Xseq or -Xraw.

For the background I got the explanation by Fons Adriaensen (10 years ago) relayed at https://wiki.linuxaudio.org/faq/start#qwhat_is_the_difference_between_jack-midi_and_alsa-midi

Then on the same wiki I also got this "interesting" comment:
"Interestingly, JACK developers (by way of Paul Davis) recommend using a2jmidid for exposing raw midi ports (https://community.ardour.org/node/4336), while the a2jmidid author recommends using the '-X raw' flag for JACK (https://gna.org/bugs/index.php?23272)."

That made me wonder what I should do. gna.org is down so I couldn’t check/understand the second comment.

I tried to locate the current repo of a2jmidid but could only find this mirror (?) referenced by the Arch packagers: http://repo.or.cz/a2jmidid.git There it says:
"Main goal of this project is to ease usage of legacy, not JACK-ified apps, in a JACK MIDI enabled system."

If I only use Ardour, Yoshimi, Qmidiarp and mididings (all of which can do jack-midi on their own), should I still connect them through a2jmidid just for the sake of timing? 
I’d be ready to join the Jack1 rebels if this could simplify my midi setup. I only need one server on one local computer, no dbus etc. But I also need low latency for live performances and some dsp overhead. On these topics I’m confronted with another decade of contradictory informations
The feature comparison on the jack website [1] seems to give a slight advantage to jack2 in my case, mainly because of "Supports multiple processors (SMP)" (I do have parallel subgraphs in my setup). That had prompted me to switch to jack2 some 3 or 4 years ago. Now the state of midi support makes me look back…

So I wonder:
- if I use jack2 and a2jmidid, what -X option should I give to  jackd at startup? Will that reproduce the native behavior  of jack1?
- are there recent benchmarks to compare jack1 vs jack2 regarding  latency and dsp load?

Cheers,
Victor

[1] https://github.com/jackaudio/jackaudio.github.com/wiki/Q_difference_jack1_jack2