URL-specific operations about mint HOT 7 CLOSED

I agree. Please open a separate issue. I think having a separate library that handles IDNA would be ideal, we would only need some documentation on how to use the the library with Mint.

from mint.

I think this HTTP library should focus on HTTP handling, rather than having all convenient features for HTTP clients.

For "url helper methods", such as validation, auto-fix:

• "modern" browser behavior is not necessarily "standard"
• If bad value is given, then HTTP library should fail, instead of trying to auto-fix it

Those "convenience" features should be built on higher level, not this HTTP library level. For example, I'm using tesla and I can do all with Tesla middleware (e.g. see Tesla.Middleware.BaseUrl) - so that underlying HTTP library (e.g. hackney, httpc..) doesn't need to handle this.

For IDNA:

We may end up with common interface for IDNA (like we do for Json with jason and poison) so that we can evolve/release mint without being coupled with specific idna library.

from mint.

Forgive my ignorance but I am not an expert in this area. What would be missing from the Elixir URI module in the stdlib that we might want to add here?

from mint.

Well, I'm not sure if it's needed in Elixir Core, but for HTTP Client next functionality might be useful:

1. URL normalization:

   # all examples are working in modern browsers
   "https://////www.google.com" -> "https:/www.google.com"
   "https://你好你好" -> "https://xn--6qqa088eba/"
   "https://０Ｘｃ０．０２５０．０１" -> "https://192.168.0.1/"
   

2. URL validation:

   https://www .google.com -> false
   

3. Relative URLs:

   "../foo?a=bar#bazz" with base "https://www.google.com/maps" -> "https://www.google.com/foo?a=bar#bazz"
   

from mint.

Thanks for starting this discussion. Since xhttp is still being developed we should do this in steps, preferably small steps.

I think we can start by adding idna. We can use the library hackney uses but in the future I would like to investigate if we can reduce the size of the idna modules by implementing them in elixir and do the same code size optimizations elixir core does for the unicode modules (joining similar clauses and so on).

I would like to wait with the other whatwg normalizations, until we have looked at it more, for two main reasons; the whatwg spec looks very complex and I'm not sure a generic http client should have the same behavior as a browser.

Does this sound good to you?

from mint.

@ericmj This sounds really good. We will pend with investigations until something will become more clear. Hope this thread will help with coordinations upon given topic in the future.

from mint.

I agree with @chulkilee. I think we can close this issue and open one that will remind us to look into a IDNA interface. Wdyt @ericmj?

from mint.