CardDAV write-back: a full troubleshooting page, and correct what the work made wrong
The four operational additions - failure reporting, the write log, D015 and
D016 - were compressed into two sections of the internals page. That page is
written for somebody changing the code; the reader who needs this is somebody
whose write-back is not working. Ed asked for them in proper depth, and he was
right that they had not had it.
New: CardDAV write-back - when it goes wrong
- the guarantee first: an analyst's edit is never lost, which is why every
message opens with "Saved here, but..."
- the asymmetry that shapes the whole design, stated as a table: a bad import
damages our copy and is recoverable, a bad write-back damages the operator's
address book and is not
- all four results, what each means, and what to do - including why Refused is
the safety net working and not a fault to be cleared
- the Digest trap on its own, with the measured numbers, because a stock Baikal
answers Basic with a 401 that looks exactly like a wrong password
- reading the log, including why the server's raw response is kept verbatim and
is the thing to forward to whoever runs the server
- D015 rung by rung, with what a failure at each rung actually implies
- D016 section by section, why there is no "fix it" button, and the trap it
reveals: drift is what makes a later write-back get refused
- a symptom-to-where-to-look table
- what is deliberately not automatic, since each is a thing people ask for
Corrected where the build had made a page wrong from a distance:
- Contact-Details still said all seven fields lock on any managed record. Now a
table of three cases, including that write-back unlocks them entirely, with
the permanently-unfillable-fields bug named rather than quietly fixed.
- System.md promised a fixed 4s toast dismiss, and its debug tools list stops at
D004 - ten tools never written up. D015 and D016 added; the gap is not
backfilled but the page now says so and names the registry as the only source
of truth. A list that looks complete and is not is worse than one that admits
it.