summaryrefslogtreecommitdiff
path: root/doc/devblog/day_201__chunky.mdwn
diff options
context:
space:
mode:
authorGravatar Joey Hess <joey@kitenet.net>2014-07-23 17:55:44 -0400
committerGravatar Joey Hess <joey@kitenet.net>2014-07-23 17:55:44 -0400
commitd0a0145fc19cd1857fbe544ca53c9cf397424ce1 (patch)
tree4e5109a0109a0380be3fa01f91ea0dc41ea9c3ad /doc/devblog/day_201__chunky.mdwn
parentf9ad0ce0524fc842850e93cb253df432ce829ed7 (diff)
devblog
Diffstat (limited to 'doc/devblog/day_201__chunky.mdwn')
-rw-r--r--doc/devblog/day_201__chunky.mdwn12
1 files changed, 12 insertions, 0 deletions
diff --git a/doc/devblog/day_201__chunky.mdwn b/doc/devblog/day_201__chunky.mdwn
new file mode 100644
index 000000000..b210702a4
--- /dev/null
+++ b/doc/devblog/day_201__chunky.mdwn
@@ -0,0 +1,12 @@
+Working on designs for better [[chunking|design/assistant/chunks]]. Having
+a hard time finding a way to totally obscure file sizes, but otherwise a
+good design seems to be coming together. I particularly like that the new
+design puts the chunk count in the Key (which is then encrypted for special
+remotes, rather than having it be some special extension.
+
+While thinking through chunking, I realized that the current chunking
+method can fail if two repositories have different chunksize settings for
+the same special remote and both upload the same key at the same time.
+Arn't races fun? The new design will eliminate this problem; in the
+meantime updated the docs to recommend never changing a remote's
+chunksize setting.