summaryrefslogtreecommitdiffhomepage
path: root/doc
diff options
context:
space:
mode:
authorJoey Hess <joeyh@joeyh.name>2017-05-20 15:16:40 -0400
committerJoey Hess <joeyh@joeyh.name>2017-05-20 15:16:40 -0400
commit34b0151e125a6698f57ea476ccfa922c6275edf1 (patch)
treec4c7f57421ae1e7b87bb0f8b82ff97a1cec93222 /doc
parent2e16195d151d401a664fa929604413aa613aa9f5 (diff)
downloaddebug-me-34b0151e125a6698f57ea476ccfa922c6275edf1.tar.gz
move unsafe hashing out of instance to avoid misuse
Avoids breaking backwards compat and should avoid future foot-shooting.
Diffstat (limited to 'doc')
-rw-r--r--doc/protocol/comment_4_6c6cd957b3e4db5b77f87b13c4e35e6b._comment35
1 files changed, 35 insertions, 0 deletions
diff --git a/doc/protocol/comment_4_6c6cd957b3e4db5b77f87b13c4e35e6b._comment b/doc/protocol/comment_4_6c6cd957b3e4db5b77f87b13c4e35e6b._comment
new file mode 100644
index 0000000..ed1bb32
--- /dev/null
+++ b/doc/protocol/comment_4_6c6cd957b3e4db5b77f87b13c4e35e6b._comment
@@ -0,0 +1,35 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 4"""
+ date="2017-05-20T17:53:29Z"
+ content="""
+So the problem comes from the hash
+"cf83e1357eefb8bdf1542850d66d8007d620e4050b5715dc83f4a921d36ce9ce47d0d13c5d85f2b0ff8318d2877eec2f63b931bd47417a81a538327af927da3e",
+-- if that's intended to be a `Maybe Hash` that's the hash of a `ByteString`,
+we can't tell if it was produced by hashing `Nothing`, or hashing
+`Just (mempty :: ByteString)`
+
+Double hashing would avoid this ambiguity, but it does also break backwards
+compatability of the debug-me protocol and logs. It's still early enough to
+perhaps do that without a great deal of bother, but it's not desirable.
+
+debug-me does not appear to be actually affected by this currently. The only
+`Maybe Hash` in debug-me is used for a hash of values of type `Activity`
+and `Entered`, not the hash of a `ByteString`. So, as far as the debug-me
+protocol goes, the above hash value is unambiguously the hash of `Nothing`;
+there's no `Activity` or `Entered` that hashes to that value.
+(Barring of course, a cryptographic hash collision which would need SHA2
+to be broken to be exploited.)
+
+So, I'd like to clean this up, to avoid any problems creeping in if
+a `Maybe Hash` got used for the hash of a `ByteString`. But, I don't feel
+it's worth breaking backwards compatibility for.
+
+(I tried adding a phantom type to Hash, so the instance could be only
+for `Maybe (Hash Activity)`, but quickly ran into several complications.)
+
+What I've done is fixed the instance to work like you suggested,
+but kept the old function as `hashOfMaybeUnsafe` and used it where
+necessary. This way, anything new will use the fixed instance and we don't
+break back-compat.
+"""]]