20:33 <GH> [Principia] scimas commented on issue #1565: There is another related bug to this (or maybe they have a common cause). Manoeuvre node isn't updated by principia if flight plan window is not open (even if show on navball was selected before closing the window).... https://git.io/vb6nn

21:57 <GH> [Principia] pleroy pushed 1 new commit to master: https://git.io/vb6lM

21:57 <GH> Principia/master 00950f0 Pascal Leroy: Christoffel

22:47 <egg> lamont: thanks for the reply on the fora, it's good to deconfuse users :-)

23:22 <xShadowx> ......egg....thinking good to be 'less' confusing? but egg spreads confusedness wherever he goes

23:36 <egg> :D

23:43 <egg> lamont: "impulsive station keeping once per orbit" we really don't support impulsive stuff because it breaks a lot of assumptions (but admittedly short bursts would only hit performance on rare occasions). For orbital decay otoh, you really need to be in the force computation all the time

23:45 <egg> which means you need to be principia because no way in hell you're serializing where you are in the atmosphere and whatnot to get the decay drag

23:47 <lamont> yeah if you’ve got internal state for the orbit you’d impulsively update it and then flush that state and start over, once per orbit or something (or possibly some other reasonable cycle at high warp)

23:47 <bees> devil's advocate - pre-compute drag profile, save it, and process it as a gravitational disturbance, bothering with forces/heat makes no sense because if they start to be a problem - you are hitting real atmo already

23:48 <lamont> and only for gravityless ship objects

23:48 <egg> yes of course you need to precompute something, but still, that "disturbance" has to be computed in the inner loop

23:49 <egg> it's not a gravitational disturbance, there's no way you're modeling that remotely decently by perturbing the geopotential field

23:49 <egg> lamont: "once per orbit" *really* isn't a thing here

23:50 <egg> lamont: and behaviour that depends on timewarp is something we try really hard to avoid

23:50 <lamont> yeah

23:50 <egg> lamont: what would have to happen is "algorithm decides when to apply correction burn" -> "correction gets applied there" (doesn't matter that it's not impulsive, you can do a tiny integration anyway, it's unfrequent so cheap)

23:51 <lamont> right

23:52 <egg> (you flush the state in any case because you change integrators in the process, the integrator for engines is a variable step RKN, not a symmetric linear multistep method)

23:52 <lamont> and i know you don’t really have an ‘orbit’ but that’s part of that last qualification. do it once per period of the keplerian tangent orbit, and if the user is trying to use station keeping on something where that approximation is a pile of failure then the station keeping will be a pile of failure...

23:53 <lamont> and for decay to be applicable you should be somewhere in the vicinity of LEO

23:53 <egg> yeah the thing is the osculating orbit is *very* shit, you'd have to have some sort of mean orbit that takes J2 into account otherwise you're just throwing fuel out the window

23:53 <egg> and yeah decay needs screening, applying jovian decay in LEO seems unadvisable

23:53 <lamont> right well that’s all my /other/ qualifications about getting the station keeping algorithm right to be patient

23:53 <egg> (we have to apply jovian forces, but that's another problem)

23:54 <lamont> right

23:54 <lamont> lol

23:54 <lamont> so yeah TL;DR its hideously complicated when you start peeling it apart

23:54 <egg> (the only way to ignore jovian forces is to integrate in a reference frame that falls towards jool)

23:54 <egg> (jupiter)

23:54 <lamont> yeah we had that discussion before

23:55 <lamont> N-body has to be N-body

23:55 <egg> if you integrate in the geocentric frame you can ignore all but the earth quite well (and integrate differences where needed)

23:55 <lamont> can’t be X here and Y there

23:55 <egg> well, you can, but it requires care and is dangerous

23:55 <lamont> yeah

23:55 <egg> that's the KSP thing, you're never building things like in real life for a set mission profile

23:55 <lamont> reference frames

23:55 <egg> it has to work for sounding rockets and Voyager

23:55 <egg> and that is *hard*

23:56 <lamont> yeppers

23:56 <egg> lamont: speaking of which, how is PEG doing

23:56 <egg> did you add J2 support?

23:56 <lamont> it is looking lots better

23:56 <lamont> no

23:56 <lamont> unfortunately not yet

23:56 <egg> (we can probably give you a stable call to get J2 tbh)

23:57 <lamont> plain old J2=0 has been pissing me off for months until like literally this week

23:57 <egg> :D

23:58 <egg> (we'd need to define a stability policy in number of synodic months of notice for deprecation if we start giving you entry points, but that should be workable :-p)

23:58 <egg> stabbity policy

23:58 <lamont> hahahah

23:59 <egg> "this function is deprecated and will be removed at the next partial eclipse"

23:59 <UmbralRaptor> Policy includes not using a reference frame that crosses Jool's event horizon in finite time.