You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Stop pages fattening to the cap by moving the split trigger off the cap itself: a soft watermark
that forces a split-or-justify decision while a seam is still cheap to find, plus a delta-synthesis
bias toward creating a new page rather than growing an existing one.
Meta
feature: docs/features/concept-store-capacity
estimate: M
candidate stories: A page crossing the watermark forces a split-or-justify decision; New
behaviour prefers a new page over a bigger one; The store is re-curated to the width the schema
assumed
This is page hygiene, and it is real: the schema's stated growth model is that the store grows by
adding files, but what actually happened is that pages fattened to the cap, because an update
delta is always cheaper to author than a create (which owes a slug decision, a domain filing and
the taxonomy gate). Nothing pushes back until 400, by which point the split is expensive and the
seam is unclear. Median own-content is 321 words against a 400 cap.
It is not the fix for the saturation raised in PR #219, and was analysed and set aside on that
basis. Splitting a high-degree page does not reduce edge prose — it increases it. Both halves keep
most of the parent's neighbours plus an edge to each other, and every added edge fires the
reciprocity fan-out onto pages that are already full. Applying this to relieve cap pressure would
have deepened the problem it was meant to solve.
Plan it as what it is — keeping pages narrow enough to stay independently loadable — and only
after #220 has removed the incentive to conflate the two problems.
Measured basis at filing
47 active pages · own-content median 321 words, maximum 383 · 13 pages at exactly 400 total body
words · 22 of 47 pages at the 7-invariant cap.
Stop pages fattening to the cap by moving the split trigger off the cap itself: a soft watermark
that forces a split-or-justify decision while a seam is still cheap to find, plus a delta-synthesis
bias toward creating a new page rather than growing an existing one.
Meta
behaviour prefers a new page over a bigger one; The store is re-curated to the width the schema
assumed
Why this is deferred, not rejected
This is page hygiene, and it is real: the schema's stated growth model is that the store grows by
adding files, but what actually happened is that pages fattened to the cap, because an
updatedelta is always cheaper to author than a
create(which owes a slug decision, a domain filing andthe taxonomy gate). Nothing pushes back until 400, by which point the split is expensive and the
seam is unclear. Median own-content is 321 words against a 400 cap.
It is not the fix for the saturation raised in PR #219, and was analysed and set aside on that
basis. Splitting a high-degree page does not reduce edge prose — it increases it. Both halves keep
most of the parent's neighbours plus an edge to each other, and every added edge fires the
reciprocity fan-out onto pages that are already full. Applying this to relieve cap pressure would
have deepened the problem it was meant to solve.
Plan it as what it is — keeping pages narrow enough to stay independently loadable — and only
after #220 has removed the incentive to conflate the two problems.
Measured basis at filing
47 active pages · own-content median 321 words, maximum 383 · 13 pages at exactly 400 total body
words · 22 of 47 pages at the 7-invariant cap.