Re: [apps-discuss] Scope of RFC3986 and successor - what is a URI?

[Sam Ruby <[e-mail]>]

On 01/17/2015 08:08 PM, [e-mail] wrote:
>
> *From:* Sam Ruby <mailto:[e-mail]>
>
>  > I have trouble with the words "existing *spec-compliant*
> implementations".
>
>  > Either that term is meaningless, or I have yet to find one.
>
>  >I've taken a look at a lot of libraries that produce and consume URIs
>  >(some call them URLs, but lets not worry about that for the moment).  I
>  >have yet to find one that is spec compliant.
>
> Again I can only speak for the tests that you are doing on the csharp
> implementation, but the test results are far from conclusive when it
> comes to comparing against RFC 3986.  From looking at your makefile, I’m
> making a guess that you are using mono to build the csharp code.  The
> System.Uri class is actually part of the .net framework.  Currently
> there are multiple versions/editions of the .net framework in addition
> to the mono version.  Even within the Microsoft .net framework there are
> configuration settings that control URI parsing behavior

I tried Mono at first, but saw very poor results.  I then removed it 
from the makefile, and reran with Microsoft .Net and saw much improved 
results.  You can see the version I ran with here:

https://github.com/webspecs/url/blob/develop/evaluate/useragent-results/csharp

> You can turn on IRI compliant parsing with a global setting.
> http://msdn.microsoft.com/en-us/library/bb882600(v=vs.110).aspx  The
> default of this global setting was changed from false to true on the
> move from .Net 4.0 to 4.5.
>
> You can also configure all kinds of other behavior related to URI parsing
> http://msdn.microsoft.com/en-us/library/bb882619%28v=vs.110%29.aspx

If you have new results or updates to the code to suggest, I'm 
interested.  My preference is GitHub pull requests, but whatever works 
for you is fine with me.

> My reason for bringing up these details is because I believe that
> declaring RFC 3986 invalid because implementations don't match its
> behavior is self-defeating.  IMHO, a spec should be a goal that
> implementations aim for, not a reflection of the behavior of an
> implementation that currently has the biggest market share.

I, too, believe that a spec should be a goal that implementations aim 
for.  I also believe that spec writers should listed to implementers.  I 
am NOT saying that that wasn't done here, but I am saying that there is 
clear evidence that a decade after RFC 3986 was published that 
implementors (with the possible exception of Microsoft?) have not 
adopted RFC 3986 as a goal.  This includes quite a number of 
implementations that make claims of compatibility.

I will say that in the case of Microsoft, they have identified the URL 
specification as one that they are considering:

https://status.modern.ie/urlapi

They have a developer who is working on this, and he has posted feedback 
to the W3C mailing list.

> Having said all this, I will try and run your tests on various versions
> of Microsoft’s .net framework and see how much the results deviate.  I
> did try running your C# tests, but they are based on a JSON file that I
> believe gets created by some javascript that I didn’t find.  My js
> skills are less than stellar!

I'll make it easier for you.  You can find the json file you need here:

https://url.spec.whatwg.org/interop/urltestdata.json

> Darrel

- Sam Ruby