unqualified attributes should have an empty namespace

[ma314smith]

I was getting an incorrect digest value when trying to validate WS-Fed metadata. It ended up being an issue with the ordering of unqualified attributes.

Expected Canonical Output:

<EntityDescriptor xmlns="urn:oasis:names:tc:SAML:2.0:metadata" ID="_8d1dcc18-2f1e-4a93-850b-e3a3081b3ca1" entityID="https://sts.windows.net/8bd6e98d-e212-4022-b13f-a244fab4c253/">
  <RoleDescriptor xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" protocolSupportEnumeration="http://docs.oasis-open.org/wsfed/federation/200706" xsi:type="fed:SecurityTokenServiceType">
  </RoleDescriptor>
</EntityDescriptor>  

Incorrect Canonical Output (note the change in order for "protocolSupportEnumeration"):

<EntityDescriptor xmlns="urn:oasis:names:tc:SAML:2.0:metadata" ID="_8d1dcc18-2f1e-4a93-850b-e3a3081b3ca1" entityID="https://sts.windows.net/8bd6e98d-e212-4022-b13f-a244fab4c253/">
  <RoleDescriptor xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance "xsi:type="fed:SecurityTokenServiceType" protocolSupportEnumeration="http://docs.oasis-open.org/wsfed/federation/200706">
  </RoleDescriptor>
</EntityDescriptor>  

This was happening because the default namespace was being added to unqualified attributes in exclusive-canonicalization.js:

if (!attr.namespaceURI && defaultNS) { attr.namespaceURI = defaultNS; }

According to the spec, unqualified attributes should have an empty namespace. See the example, here and the note about "attr2": http://www.w3.org/TR/xml-c14n#Example-SETags

This pull request removes the line that was adding the default namespace, and includes a unit test to cover this scenario.

Note that 2 of the unit tests are failing for me (same behavior on Windows and on http://www.tutorialspoint.com/codingground.htm). The failing tests are "test with a document" and "test validating SAML response". I've verified that they were also failing without my change, so it doesn't appear to be related. They both use the same test xml, and a quick look at the canonical output shows some random 
 values, so maybe there are some whitespace/line endings that aren't being handled properly.

[bjrmatos]

seems fine to me.. all tests passed in OSX, i will test in windows to see if i get failing tests

[bjrmatos]

i made some changes in order to normalize line endings across OS, could you fix your PR and re-run the tests?.. also let me know if you still see the 2 failing tests in windows, i will test on windows later in order to verify the test cases.

[ma314smith]

I'm still seeing the same 2 tests failing.

[bjrmatos]

all test passed for me in windows 7 and node 0.10.36 (in current xml-crypto master branch)

could you delete your copy of xml-crypto (and fork), download again from github and re-run the tests?

in the past i noticed that some tests fails because git convert line endings based on the configuration found in the local computer.. however with the commit that i have done yesterday this should be resolved (I just checked this on my windows PC)

[bjrmatos]

also.. in my windows all test passed with your fork

[ma314smith]

ok, deleting my local copy and pulling it down again resolved the issue. all tests are passing now. i think we're good.

[bjrmatos]

ok, all fine for me.. @yaronn are you ok with this? according to the spec this is the correct behaviour

[yaronn]

looks good to me!

[bjrmatos]

done! thnks @ma314smith