Gleipnir : ... Haul the ends to tension the bundle, and
as the tension reaches the middle nipping loop,
the Gleipnir works its magic and grips the ends without any slippage.
One quibble with this, noting the emphasized part:
there is sometimes a *delay* or difficulty of getting
sufficient tension around some object (overcoming
frictional resistance) to achieve adequate nipping.
(And, should the bound entity --esp. pertinent where
this isn't a solid object-- change in its bound measure,
the binding if slackened can fail (a reason to tie off
the structure, in such cases, or use some other binding).)
Uber Gleipner or UG : Middle the cord and pass the bight end around the bundle to meet the two ends. Twist and fold the bight end to form a round turn. Pass the ends (or slips) through the round turn from opposite sides. Haul the ends (or the slips) to the desired tension. Done.
Quite well described,
and everybody should get it right,
from "only" these words, without further
guidance! Now, there are different crossings
of the tails, but these are fine points awaiting
elucidation as to what difference they might make.
(Further : frankly, I find your particular image much
harder to comprehend(!); I imagine a larger object,
such as a sleeping-bag/bed-roll (carpet roll would
be something needing stout resistance), and a longer
lead-in to the
turNip, and good view of the tails
oriented roughly parallel to binding axis, hauled out.)

... the improvements are arguably trivial ...
In the proposed UG variant, the tension applied to the ends applies tension directly to the round turn nipping loop, nothing is lost in having to travel around the bundle, so the grip is at its maximum. (If you want still more grip, make two round turns in the byte end).
...
What do you think?
Precisely the redress my original misgivings sought.
And I've tried other variants, with the same design
goal : to have the tensioning have immediate effect
on the
turNip ("nipping loop"), avoiding the mitigation
of force that might come from friction around what's bound.
(And Xarax presented a version in which the original's
turNip was inverted, taking the delivery up away
from the object then down-around, gaining I think
some slight but appreciable force and locking friction.)
But this revision still suffers a problem with uses
where one might be e.g. disposing of the material
with the bound contents : that of efficiency in the
consumed length (as one would need sufficient
length to enable the set-up and binding, but the
have at least one long tail lost (the other end can
be cut close and the spool of twine keeps it for the
next task)).
--dl*
====