lightweight
com.atproto.sync.listReposByCollection
todos
+6
-6
+6
-6
hacking.md
+6
-6
hacking.md
···
72
72
- [x] #sync shortcuts
73
73
- [x] noop if rev/cid are unchanged
74
74
- [-] drop any preceeding #commits from the queue? (tricky with noop)
75
+
- [x] swap in repo-stream for backfill
76
+
- [x] with memory limit
77
+
- [-] with a global concurrency limit for big repos (maybe later)
78
+
- [-] with disk spilling for huge repo (maybe later)
79
+
- [-] with queueing resync for large repos if resources are taken?? (maybe later)
80
+
- [x] (self-reminder: get_repo should be rare in lightrail)
75
81
- [ ] "deep crawl" mode for relays that listHosts -> listRepos on host instead of relying on relay listRepos
76
82
- [ ] defensive loop-cursor handling
77
83
- [~] actually firehose-index!!
···
81
87
- [ ] swap in repo-stream
82
88
- [~] prefix-merge walker (limit by total collections to be merged?)
83
89
- [x] add an all-collections index
84
-
- [ ] swap in repo-stream for backfill
85
-
- [ ] with memory limit
86
-
- [ ] with a global concurrency limit for big repos
87
-
- [ ] with disk spilling for huge repo
88
-
- [ ] with queueing resync for large repos if resources are taken??
89
-
- [ ] (self-reminder: get_repo should be rare in lightrail)
90
90
- [ ] lenient sync1.1
91
91
- [ ] *don't* allow non-validating commits that look like sync1.1
92
92
- [ ] rachet by PDS host: be lenient if we have never seen a sync1.1-looking commit, always strict after we see one.