Editorial: clarify URL validity #666
Conversation
url.bs
Outdated
| @@ -1170,7 +1170,9 @@ unified model would be, please file an issue. | |||
|
|
|||
| <li><p>The <a>URL serializer</a> takes a <a for=/>URL</a> and returns an <a>ASCII string</a>. (If | |||
| that string is then <a lt="URL parser">parsed</a>, the result will <a for=url>equal</a> the <a | |||
| for=/>URL</a> that was <a lt="URL serializer">serialized</a>.) | |||
| for=/>URL</a> that was <a lt="URL serializer">serialized</a>.) The output of the | |||
| <a>URL serializer</a> is not always a <a>valid URL string</a>. I.e., not all <a for=/>URLs</a> are | |||
There was a problem hiding this comment.
It might be worth adding a pointer to #379, because I am still hoping we can change this.
There was a problem hiding this comment.
You don't think it's useful that \ in HTTP URLs shows up as something you probably want to fix?
There was a problem hiding this comment.
I mean, we can rehash that thread here if you want... yes, I think the serializer should always produce valid URLs, either by expanding the definition of valid, or changing the serializer.
There was a problem hiding this comment.
Oh sorry, \ is not applicable then I think. I got this confused with the people who think all inputs ought to be valid or rejected. Whereas you don't necessarily think all inputs ought to be valid or rejected, but the invalid inputs that are accepted, ought to be transformed to something valid when they are spit out again.
So yeah, the reason for that is mainly encouraging RFC 3986 interop. But I'm not sure anyone is really appreciative of that.
url.bs
Outdated
| @@ -1160,7 +1160,7 @@ unified model would be, please file an issue. | |||
|
|
|||
| <ul> | |||
| <li><p>The <a>URL parser</a> takes an arbitrary string and returns either failure or a | |||
| <a for=/>URL</a>. | |||
| <a for=/>URL</a>. It might also record zero or more <a>validation errors</a>. | |||
There was a problem hiding this comment.
Do we know if
"URL parser records zero validation errors" implies "input string is a valid URL string"?
How about the other direction?
It would be great to clarify the purpose of these validation errors.
There was a problem hiding this comment.
We do not, and I strongly suspect they are not equivalent. I think there are some open issues on it.
My preferred strategy has been to instrument whatwg-url with both modes of validation and fuzz to find examples where they mismatch. I haven't made the time to do so yet though.
…roposal or consensus)
annevk
added
the
topic: validation
Closes #595.
Preview | Diff
Preview | Diff