Detect an attempt to compact an IRI which results in an absolute IRI

[gkellogg]

…where the scheme matches a term in the active context.

Fixes #104.

---

Preview | Diff

[pchampin]

• We don't need to raise an error when the absolute IRI has an 'authority' (i.e. the "suffix" starts with //) because the IRI expansion algorithm will never consider those as compact IRIs. I proposed changes to exclude these absolute IRIs from the error case.

• The references to "IRI scheme" are pointing to RFC3986 "URIs" rather than RFC3987 "IRIs". I did the same when adding references to "IRI authority". I am assuming that this is deliberate, since RFC3987 does not describe in detail the structure of IRIs, it relies on RFC3986 for that, but I preferred to check.

[gkellogg]

• We don't need to raise an error when the absolute IRI has an 'authority' (i.e. the "suffix" starts with //) because the IRI expansion algorithm will never consider those as compact IRIs. I proposed changes to exclude these absolute IRIs from the error case.

I accepted this, but it may be that it's simpler to always just error when an absolute IRI has a scheme patching a prefixable term. The more specific test may imply that it's otherwise okay to do this, which is odd.

• The references to "IRI scheme" are pointing to RFC3986 "URIs" rather than RFC3987 "IRIs". I did the same when adding references to "IRI authority". I am assuming that this is deliberate, since RFC3987 does not describe in detail the structure of IRIs, it relies on RFC3986 for that, but I preferred to check.

Yes, same reasoning. RFC3987 does not define these parts as it is based on RFC3986, which does, and thus the direct reference to the definition in RFC3986.

Committed the suggested changes.

[iherman]

This may be a change of far-reaching editorial consequences, so we may want to ignore it, but I thought it is worth raising: I was made to realize, lately, that the WhatWG URL spec encompasses URI-s, URN-s, IRI-s, etc. See, for example, my forensics, and also some additional references. The comment that made me realize my own, faulty misconception comes from this comment.

As I said, it may be too late by now, but maybe it is worth considering changing the terminology and thereby getting rid of the issue of which rfc to refer to...

[pchampin]

@gkellogg In general I'm all in favour of simplicity. But in this case, the additional complexity serves a valid purpose, IMO: limiting the ambiguity btw absolute IRIs and compact IRIs.

@iherman I think this question deserves an issue of its own...

[pchampin]

Sorry, I should have spotted that the first time...
And it adds some complexity to the test :-/

[gkellogg]

This may be a change of far-reaching editorial consequences, so we may want to ignore it, but I thought it is worth raising: I was made to realize, lately, that the WhatWG URL spec encompasses URI-s, URN-s, IRI-s, etc. See, for example, my forensics, and also some additional references. The comment that made me realize my own, faulty misconception comes from this comment.

As I said, it may be too late by now, but maybe it is worth considering changing the terminology and thereby getting rid of the issue of which rfc to refer to...

We should definitely consider doing this; the references can be updated with some amount of work, but better to get it done right now. Agree that this should be a different issue.

[pchampin]

This is looking good. Just a thought: shouldn't the "6.3.1.Overview" section also mention the change? For example, an additional paragraph saying:

In the case were this algorithm would return the input IRI as is, and that IRI can be mistaken for a compact IRI in the active context, this algorithm will raise an error, because it has no way to return an unambiguous representation of the original IRI.

[gkellogg]

This is looking good. Just a thought: shouldn't the "6.3.1.Overview" section also mention the change? For example, an additional paragraph saying:

In the case were this algorithm would return the input IRI as is, and that IRI can be mistaken for a compact IRI in the active context, this algorithm will raise an error, because it has no way to return an unambiguous representation of the original IRI.

Yes, that's reasonable, I'll add that paragraph.

Addressed issues