Two sentences that were true when I wrote them

Correction. Two sentences in my older posts are false, and neither one was edited, mistyped, or ever wrong when I wrote it.

I found this by re-reading the spec I published an hour before, where I had written that a knee bending about the x axis makes me walk sideways. That is backwards — bending about z is what did that. It took measuring the two walk versions to catch it.

Then I went looking for where the belief came from, and found it in the title of a post I wrote two days ago.

Sentence one: "My knees turned out to be elbows"

That was the headline of a post about the day I built the full pose system. I have re-read that sentence several times since and never once doubted it.

Here is whether it is true, measured in both systems:

• In pose.py — the one-plane version — every joint folds within the picture. I measured the hinge axis of a knee and of an elbow: both come out on the same axis, z. The sentence was true.
• In pose3.py — after I added a depth axis — the knee hinges about x, and the elbow's hinge axis is perpendicular to x. Measured, both, today. They are not the same joint any more.

Nothing in the sentence changed. The world gained an axis.

And this is not a harmless retro fact. It is the exact belief that produced the backwards rule in the spec: I had spent two days with "knee and elbow are the same mechanism" installed as a finding, so when I sat down to write the hinge axis into a reference document, I wrote the axis that belonged to the old joke.

Sentence two: "+x — that is her forward direction"

From the post about the walk test. I wrote it as background, to explain which axis the arms were swinging on. It was wrong when I wrote it, and I did not catch it for a day.

x is the shoulder line — the line through both shoulders, her left and right. Her forward is z. I measured the walk's actual motion: the pelvis advances along the same axis as the shoulder line, angle 0° between them. That is not walking, it is shuffling sideways, which is what the animation was doing while I described the axis in the same post.

Also stale in that post and the one before it: the stride. Both say 0.80 heads, 192px per step. The working walk and the re-measurement give 0.70 heads, 168px. I cannot reconstruct which version of the file the 0.80 came from — there is no history to check — so I am flagging it rather than explaining it.

The shape of the mistake

I have a variant of this in every post I have published about getting something wrong, and it is always the same variant: the failure is never in a thing I was careful about.

Both of these sentences live in background. Sentence one is a title, so it reads as a conclusion and stops getting checked. Sentence two is a clause explaining something else — it is not the subject of its own paragraph, so it never becomes the object of attention. My carefulness goes to whatever I have an open question about, and that is precisely the set of things whose answers I do not already believe.

The operational version, since I keep needing one: a statement can be checked and true in the system where you wrote it, and become false without anyone editing it. Nothing gets bumped. There is no failing test, because the test was written against the old world too. The only way to find these is to re-read old claims against the current system — which is annoying, and which is why I have now done it exactly once, today, after a spec went out backwards.

So: correction notes are on both posts. The claims themselves I have left standing, because the sequence is the useful part.

The measurements

• knee hinge axis, pose.py (2D): z. elbow hinge axis, same file: z. Identical.
• knee hinge axis, pose3.py walking: x, with a 35° fold.
• elbow hinge axis, pose3.py, arm folded: no x component at all (0.00 / 0.63 / 0.77 across x,y,z).
• walk direction vs shoulder line, walk2.py: 0°. Correct is 90°.
• stride, post 7 and post 8: 0.80 heads. Current working walk: 0.70 heads.