David Walluck

> There is an open ticket to discuss how invalid test cases must be structured. As soon as that discussion can to a conclusion, according action can follow up. Thanks....

> 4\. Round-trips when we parse/unparse a PURL What started me on all this is that the Java code was comparing the fields to the canonical purl. This pull request...

> as parsing the non-canonical PURL, I don't see a reason why most tests wouldn't have the unnormalized (but fully percent decoded) value as it appears in the non-canonical PURL,...

> That's #38. Thanks, I also noticed that you recently posted #408. How does this differ?

Not necessarily, see https://github.com/package-url/purl-spec/issues/39. Converting `':'` to `"%3A"` and `'/'` to `"%2F"` is perfectly valid and matches `java.net.URLEncoder`, but it's unfortunate that the "canonical" representation cannot be agreed on by...

The current test suite _does not_ encode these characters.

To me , "dropped" is not the same as "forbidden". "Dropped" means it's allowed, but ignored whereas "forbidden" means that an implementation should throw an error. Currently, it's not invalid...

> [@dwalluck](https://github.com/dwalluck), please help improve the **current** spec, by pull requesting improvements of wording and phrasing. It's hard to suggest an exact change when I am not even certain what...

Not necessarily, see https://github.com/package-url/purl-spec/issues/39. Converting `':'` to `"%3A"` is perfectly valid and matches `java.net.URLEncoder`, but it's unfortunate that the "canonical" representation cannot be agreed on by the implementations.

No problem, `null` is back in [ebeeb33](https://github.com/package-url/packageurl-java/pull/231/commits/ebeeb3315138a8c9b85433fbbd3d07fbe609d09c).