URL Encoding

Theory

When fetching resources, URL encoding and normalization is in theory a non-issue, since you're supposed to just reuse the same exact strings that the server gave you. For comparing URIs, see section 6 of the RFC.

Bugs

• ownCloud Contacts doubly-encodes/quotes hrefs in addressbook-multiget responses (%2540 represents @), but PROPFIND responses quote @s only once (%40). Most requests accept both @ and %40. https://github.com/owncloud/contacts/issues/581

• URLs sent to Radicale are unquoted in the response body. https://github.com/Kozea/Radicale/issues/298

Workarounds

• It is strongly recommended to avoid unsafe characters in names when PUT-ing new items or when creating collections. Safe characters include alphanumeric characters and _.-.

• For normalization, one may decode the URL until:

  • It doesn't change anymore.
  • A remaining %-sequence would decode to a reserved character.

  then encode it once. Note that this might change the meaning of the URL and accidentally give you a reference to a different item or collection, but given the current implementations and their buggyness, this is rather unlikely.

Further references

• http://lists.calconnect.org/pipermail/caldeveloper-l/2015-July/000893.html