continuum - P2P replicated document storage service

	Commit message (Collapse)	Author	Age	Files	Lines
*	HyParView (slim)	tv	20 hours	1	-2/+7
\|
*	plumtree	tv	23 hours	1	-0/+12
\|
*	indirect probes	tv	25 hours	1	-0/+6
\|
*	unlisted members forgotton	tv	33 hours	1	-6/+12
\|
*	handshakes counted	tv	33 hours	1	-0/+7
\|
*	working links recorded	tv	34 hours	1	-0/+1
\|
*	heads heard in announcements	tv	35 hours	2	-3/+5
\|
*	heads heard in replies	tv	35 hours	2	-1/+4
\|
*	heads heard forget sent	tv	35 hours	2	-1/+31
\|
*	worker sends off the pass	tv	35 hours	1	-14/+43
\|
*	suspend detected	tv	35 hours	1	-0/+4
\|
*	rtt table pruned	tv	35 hours	1	-0/+1
\|
*	one grace wait per peer	tv	35 hours	1	-1/+3
\|
*	reach sets bounded	tv	35 hours	1	-5/+11
\|
*	strangers messages small	tv	35 hours	1	-0/+4
\|
*	lanes comment	tv	48 hours	1	-2/+5
\|
*	round trips in trees	tv	2 days	1	-1/+8
\|
*	suspect paths	tv	2 days	1	-0/+5
\|
*	deflate blob chunks	tv	2 days	1	-1/+2
\|
*	epoch tag in sync	tv	2 days	1	-3/+21
\|
*	time blob messages	tv	3 days	1	-0/+19
\|
*	slow round back off	tv	3 days	1	-0/+5
\|
*	hello again when heard from	tv	3 days	2	-2/+19
\| \| \| \| \| \|	While we wait for a peer's HELLO_REPLY, anything else that comes from it makes our HELLO go again at once, at most once a second, instead of when the backoff says.
*	unanswered pull frees the slot	tv	3 days	1	-0/+1
\| \| \| \| \| \|	When the answer to a pull is overdue (the sync reply timeout: 1 s in the simtests, 10 s in production), the pull gives up the slot, and the next waiting neighbour is pulled at once.
*	overdue sync asked after from then	tv	3 days	1	-3/+3
\|
*	pull after grace	tv	3 days	1	-1/+1
\| \| \| \| \| \| \| \|	Before pulling, a node waits one measured round trip to the announcer (1 s at most) and pulls only what is still missing. If the round trip isn't known yet, it pulls at once, as now. On the clean benchmark links the round trip is a few milliseconds, so those scenarios shouldn't change.
*	announcements answered when lacking	tv	3 days	2	-5/+23
\| \| \| \| \| \| \|	A neighbour now answers an announcement only if it lacks some of the announced heads. Heads that merely differ are no longer enough, since they do while writes go on. The check reads the document, so it runs off the message consumer, as the pull check already does.
*	unanswered members not reached	tv	4 days	1	-7/+18
\|
*	replace eager links by trees	tv	4 days	1	-8/+3
\|
*	measure 2-hop neighbourhood	tv	4 days	2	-3/+11
\|
*	introduce eager links	tv	5 days	1	-3/+8
\|
*	skip neighbours with our heads	tv	5 days	1	-0/+1
\|
*	announcements answered only if heads differ	tv	5 days	3	-2/+24
\|
*	retry new peers before probing	tv	5 days	1	-1/+1
\|
*	probe list members	tv	5 days	2	-6/+7
\|
*	add per-node traffic counts	tv	5 days	1	-1/+9
\|
*	probe when reachable	tv	6 days	2	-1/+26
\|
*	probe wakes peer worker	tv	6 days	1	-2/+3
\|
*	peer worker forget records on change	tv	6 days	3	-3/+17
\|
*	peer worker sleeps until deadline, not fixed time	tv	6 days	4	-14/+80
\|
*	state: drop epoch_tx	tv	6 days	1	-1/+2
\| \| \| \|	The stamp is the only lock-free epoch label.
*	rename AutoMerge to Document	tv	7 days	1	-2/+2
\|
*	add peer clock skew warning	tv	7 days	1	-0/+1
\|
*	pong carries time	tv	7 days	2	-4/+7
\|
*	dashboard: try to get links dormant immediately	tv	7 days	1	-6/+12
\|
*	centralize call to topology_change	tv	7 days	4	-27/+23
\|
*	unify request_sync logic	tv	7 days	4	-16/+4
\|
*	unify heard_from logic	tv	7 days	2	-23/+35
\|
*	unify closing logic	tv	7 days	1	-3/+3
\|
*	unify sync_on predicate	tv	7 days	1	-7/+2
\|