c       The order of the elements in the file is important, and for several
c       of the elements, the SIGN of the element is important.
c       These conventions have been adopted so that the principle form
c       resembles the Astronomical Almanac (Sec. E).  Entering elements with
c       negative signs facilitates inclusion of bodies whose elements are
c       given in IAU Circulars.  The order is given below.
c       '{P|C|A}Name' {a | -q} e i Node {wpi | -w} {M | -L} epoch repoch
c       The quantities are defined as follows.
c                  P(lanet), C(omet), A(steroid)
c            a:    orbital semi-major axis OR mean distance [AU]
c           -q:    perihelion distance [AU]
c            e:    orbital eccentricity
c            i:    inclination of orbit to ecliptic [degrees]
c         Node:    longitude of ascending node on the ecliptic [degrees]
c           wpi:    longitude of perihelion (Ver. Eq.->Node->Perihelion) [degrees]
c           -w:    argument of perihelion (w = wpi - Node) [degrees]
c            M:    mean anomaly at epoch (M = L - wpi) [degrees]
c           -L:    mean longitude at epoch (L = M + wpi) [degrees]
c        epoch:    date at which [L | M] is valid (osculation date) [JD]
c       repoch:    precession epoch for i, Node, {wpi | w} [JD]
c       If q is specified, then the mean longitude/anomaly is ignored and
c       the osculation date (epoch) is presumed to be the perihelion date.
c       Only occassionally will a cometary orbit have an osculation date
c       different than the perihelion date.
c       Upon being read in, all elements are precessed to depoch.

so, taking the cue from what I had working

Constraints in all cases
0 <  a
0 <= q
0 <= e
0 <= i    < 180
0 <= Node < 360
0 <= wpi  < 720
0 <= w    < 360
0 <= M    < 360
0 <= L    < 360

This gives two ways to specify an orbit

For elliptical orbit (specified by "a")
a e i Node (wpi | w) ( M | L) osculationEpoch coordEpoch
8 parameters, two are pairs of alternatives
additional constraint:  e < 1

For orbit with any eccentricity (specified by "q")
q e i Node (wpi | w) perihelionEpoch coordEpoch
7 parameters, one is pair of alternatives

all Epoch in time scale TDB =~ TT (difference irrelevant)

I believe that parameters a q e should have those names
I am open to other names for other params, with likely ones
being something like
Node = Node
w    = argPerihelion                    (used in IAUC,MPEC)
i    = Inclination
wpi  = longPerihelion                   (used in Astronomical Almanac)
M    = meanAnomaly                      (used in IAUC,MPEC)
L    = meanLongitude                    (used in Astronomical Almanac)

The constraints and names here are not appropriate for an ephemeris
for a binary extrasolar object.  I do not know if reports of such
things are common enough that VOEvent would regret being broad enough
to accommodate them.

%%

Okay, just for completeness.

Binary star orbits have these different constraints.

-90 <= i    < 90        only known if there is astrometry and spectroscopy
0   <= Node < 180       measured from N toward E on sky

I don't think wpi (longPerhelion) is ever used for binary stars.
Astrometric binary stars have a in arcsec, not AU.
Spectroscopic binary stars have a in km.
Any binary star also needs P (period), it's not redundant there.
Most binary stars will use epoch of periapsis rather than M or L
So that's

a e i Node w periapsisEpoch coordEpoch P