continuum - P2P replicated document storage service

	Commit message (Collapse)	Author	Age	Files	Lines
*	api: /stream & dashboard send X-Accel-Buffering	tv	37 hours	3	-1/+17
\|
*	api: allow auth delegation	tv	37 hours	3	-28/+196
\|
*	dashboard: shorten long figures title	tv	38 hours	1	-0/+6
\|
*	dashboard: add B/s	tv	38 hours	1	-21/+44
\|
*	dashboard: add msg/s	tv	38 hours	1	-1/+53
\|
*	dashboard: message dot speed is configurable	tv	39 hours	1	-16/+37
\|
*	dashboard: add console filter	tv	39 hours	1	-2/+43
\|
*	dashboard: focus wins over hover	tv	42 hours	1	-22/+31
\|
*	dashboard: nodes and edges tint related messages	tv	42 hours	1	-3/+40
\|
*	dashboard: messages tint related nodes and edges	tv	42 hours	1	-5/+36
\|
*	add per-node traffic counts	tv	42 hours	3	-2/+64
\|
*	document: put user in commit message trailer	tv	47 hours	2	-8/+77
\|
*	remake: read ops, not diffs	tv	47 hours	2	-6/+6
\| \| \| \|	This gives bench_replay a speedup of ~8x/64x/280x.
*	api: /admin/changes read ops, not diffs	tv	47 hours	2	-10/+90
\| \| \| \| \| \| \| \| \| \| \| \|	The new store::changed_by reads each change's own operations instead of diffing the document before and after it. Each object's path is found once per change, as it was right after that change. So the cost now grows with the size of the change, not the size of the map it touched. - It names paths the same way the diff did: a map entry by its key, and list and text edits by the list or text. - It relies on import_obj, which Automerge marks as hidden from its docs, to turn an operation's object ID back into an object.
*	api: drop $float as JSON number is OK	tv	47 hours	3	-34/+48
\|
*	api: add type casts	tv	47 hours	5	-6/+244
\|
*	add remake benchmarks	tv	48 hours	4	-3/+94
\|
*	api: types are sticky	tv	48 hours	4	-98/+298
\|
*	api: add application/vnd.continuum.typed+json	tv	2 days	3	-189/+376
\|
*	api: add application/vnd.continuum.* scalar types	tv	2 days	5	-122/+768
\|
*	add document benchmarks	tv	2 days	2	-0/+172
\|
*	api: add x-continuum-heads to writes	tv	2 days	2	-26/+45
\|
*	probe when reachable	tv	2 days	10	-1/+69
\|
*	probe wakes peer worker	tv	2 days	1	-2/+3
\|
*	document: create new compound values at once	tv	2 days	1	-11/+29
\| \| \| \| \|	This speeds up large writes as new maps and lists are created in one batch operation instead of one operation per value.
*	document: introduce patch log	tv	2 days	5	-22/+148
\|
*	api: add support for counters	tv	2 days	3	-5/+86
\|
*	api: allow reading past versions	tv	2 days	5	-9/+308
\|
*	api: introduce /admin/changes	tv	3 days	3	-1/+141
\|
*	record node ID and API user for every write	tv	3 days	3	-14/+77
\|
*	api: PUT changes only what differs	tv	3 days	1	-23/+148
\|
*	dashboard: always try showing ping/pong alias	tv	3 days	2	-1/+30
\|
*	we're a replicated document, not KV store	tv	3 days	2	-3/+3
\|
*	ping & pong are tied by nonce only	tv	3 days	2	-5/+18
\|
*	large messages over slow links are deltas	tv	3 days	4	-14/+85
\|
*	introduce per-transport sync reply timeout	tv	3 days	9	-13/+56
\|
*	path probes follow keepalive	tv	3 days	3	-9/+41
\|
*	libp2p: idle timeout follows keepalive	tv	3 days	5	-2/+22
\|
*	tor: read max sync message	tv	3 days	1	-3/+5
\|
*	links: wait for subscriber instead of ticker	tv	3 days	1	-6/+16
\|
*	remake: download wakes on end instead of ticker	tv	3 days	2	-6/+15
\|
*	peer worker forget records on change	tv	3 days	6	-4/+32
\|
*	peer worker sleeps until deadline, not fixed time	tv	3 days	10	-32/+170
\|
*	bootstrap: replace pull by progress notify	tv	3 days	11	-2/+52
\|
*	dashboard: intended edges are loose	tv	3 days	1	-7/+14
\|
*	dashboard: allow message inspection	tv	3 days	3	-30/+287
\|
*	dashboard: 24h-clock	tv	3 days	1	-1/+4
\|
*	dashboard: one spring per edge	tv	3 days	1	-4/+6
\|
*	dashboard: figures open on press	tv	3 days	1	-2/+14
\|
*	add lost vote with restart test	tv	3 days	1	-1/+81
\|