V4 Signer

[jeskew]

The signatures in the functional tests in __tests__/SignatureV4.ts were calculated using the PHP SDK's signers. Tests use node crypto packages, with the expectation that browser SDK modules will inject the browser Sha256 implementation via the SignatureV4 constructor.

Jest reports the following in Node 4, 5, 6, 7, and 8:

Test Suites: 8 passed, 8 total
Tests:       52 passed, 52 total
Snapshots:   0 total
Time:        1.048s
Ran all test suites.
-------------------------|----------|----------|----------|----------|----------------|
File                     |  % Stmts | % Branch |  % Funcs |  % Lines |Uncovered Lines |
-------------------------|----------|----------|----------|----------|----------------|
All files                |      100 |      100 |      100 |      100 |                |
 SignatureV4.ts          |      100 |      100 |      100 |      100 |                |
 cloneRequest.ts         |      100 |      100 |      100 |      100 |                |
 constants.ts            |      100 |      100 |      100 |      100 |                |
 credentialDerivation.ts |      100 |      100 |      100 |      100 |                |
 getCanonicalHeaders.ts  |      100 |      100 |      100 |      100 |                |
 getCanonicalQuery.ts    |      100 |      100 |      100 |      100 |                |
 getPayloadHash.ts       |      100 |      100 |      100 |      100 |                |
 moveHeadersToQuery.ts   |      100 |      100 |      100 |      100 |                |
 prepareRequest.ts       |      100 |      100 |      100 |      100 |                |
-------------------------|----------|----------|----------|----------|----------------|

[chrisradek]

I had 2 main comments.

1. We talked about this yesterday but just reiterating here. It might be nice to allow customers to specify a list of unsignable headers that can be merged into our known list of unsignable headers. Ruby currently allows this, and it could be nice if they need to not sign a header that gets modified by a proxy, or we need to work-around an issue like we did with XRay.

2. Are you still expecting to have separate node and browser packages? It might be a lot to ask customers to also provide a SHA256 implementation if we can easily hook that up for them.

[chrisradek]

Nice test! Take that, frames!

[jeskew]

Turns out to be surprisingly relevant to running Jest, too.

[jeskew]

1. I've added a parameter to the object accepted by presignRequest and signRequest that allows customers to specify a custom set of unsignable headers. For performance and environment compatibility reasons, this parameter is of type {[key: string]: any}, and header names are lower-cased and then checked against this object with the in keyword.

2. Yes, definitely. At a minimum, the browser and node specific packages would supply the correct sha-256 implementation.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs and link to relevant comments in this thread.