Clean up hasher discussion on HashMap #36557
Conversation
* We never want to make guarantees about protecting against attacks.
* "True randomness" is not the right terminology to be using in this
context.
* There is significantly more nuance to the performance of SipHash than
"somewhat slow".
|
I'm not sure about "We never want to make guarantees about protecting against attacks." This has been a stated goal since Rust has had hash maps. If we aren't guaranteeing that then why bother doing it? |
|
Guarantees don't tend to exist in the security space IME. We believe SipHash 1-3 is resistant to discovery of its internal key. We believe it is hard to generate collisions without knowing that key. We believe the kernel is giving us high quality random data, etc. |
|
Ping |
| /// require this behavior you can create your own hashing function using | ||
| /// [BuildHasherDefault](../hash/struct.BuildHasherDefault.html). | ||
| /// By default, `HashMap` uses a hashing algorithm selected to provide | ||
| /// resistance against HashDoS attacks. The algorithm is randomly seeded, and a |
There was a problem hiding this comment.
Is there perhaps a wikipedia article or something we could link to here with "HashDoS attacks" ?
There was a problem hiding this comment.
I wasn't able to find anything that seems like a particularly good source to link to, unfortunately.
| /// The hashing algorithm can be replaced on a per-`HashMap` basis using the | ||
| /// `HashMap::default`, `HashMap::with_hasher`, and | ||
| /// `HashMap::with_capacity_and_hasher` methods. Many alternative algorithms | ||
| /// are available on crates.io. |
There was a problem hiding this comment.
Perhaps this could mention "such as fnv" which seems to be the most popular alternative?
|
@bors: r+ |
|
📌 Commit aaf32aa has been approved by |
|
⌛ Testing commit aaf32aa with merge f390451... |
|
@bors: retry force clean
|
Clean up hasher discussion on HashMap
* We never want to make guarantees about protecting against attacks.
* "True randomness" is not the right terminology to be using in this
context.
* There is significantly more nuance to the performance of SipHash than
"somewhat slow".
r? @steveklabnik
Follow up to discussion on #35371
context.
"somewhat slow".
r? @steveklabnik
Follow up to discussion on #35371