1.4 Beta builds broke RSS feeds with specific characters

[Asinin3]

In 1.3.1, a RSS feed such as: https://nyaa.si/?page=rss&c=1_2&f=1&q=%title%+-[SubsPlease] Would work just fine in hiding results from the specified group. However, 1.4 Beta builds broke this functionality. Specifically the: +-[SubsPlease] was what caused the problem. And the way to resolve it was by encoding that section of the url so it would turn into: %2B-%5BSubsPlease%5D Which would be used like this: https://nyaa.si/?page=rss&c=1_2&f=1&q=%title%%2B-%5BSubsPlease%5D

I'm guessing something changed internally and Taiga no-longer encodes the RSS url itself.

[erengy]

It's because v1.4 includes a new HTTP library that conforms to RFC 3986. [ and ] are reserved characters that are not allowed in the query component of a URL. Previous versions worked only by chance by naively decoding and re-encoding the components while parsing a URL string.

The subject is actually quite complicated, as there is no single definition of what a URL is. AFAIK most implementations try to conform to RFC 3986, while web browsers are overall more lenient in parsing. WHATWG's URL standard, along with a bunch of other "living standards", describe what some web browsers do. From what I understand, they don't apply well to other applications.

[Asinin3]

Can taiga automatically encode the URL instead then?

[wopian]

As per RFC 3986 2.1 and 6.2.2.2, HTTP libraries must percent encode URI component octets whose character is outside of the allowed character sets.

So to fully conform to 3986, [ and ] must always be encoded by the HTTP library (as browsers do too). Their percent encoded equivalents are perfectly valid characters and have widespread usage in APIs

[Asinin3]

What I mean is, can the field that you type in be normal plaintext, but then in the background Taiga just encodes it. It would make it far simpler for users.