Currently segments are basically triple url encoded:

• each value in a condition is encoded once to allow segment operator characters
• each value is encoded again, because the code server side urldecodes the segment condition (so the value is double encoded at this point)
• and the entire segment is encoded in the URL and matomo uses this value, not the urldecoded one in $_GET/$_POST (so the value is triple encoded at this point)

All three urldecodes happen in PHP when processing a Segment. The system works, but is overly complicated and in some cases, can introduce subtle bugs. For example, based on whether the correct amount of encoding is used or not in a segment string, the hash will change, which means it's possible to use equivalent segments but get no data from the API.

We could fix this by using a different encoding scheme to encode segment condition values, and then only urlencoding the segment in the URL. For example, we could urlencode values then replace % characters w/ $ as is done in datatable row action parameters. Then in PHP we would use the segment value found in $_GET/$_POST w/o issue.

Backwards Compatibility

To do this w/o breaking BC, we'd have to:

• still support old triple urlencoded segment values in the URL (preferrably by converting these segment parameters to the proper encoding)
• and still support looking for triple urlencoded segment archives (where the hash will be different than w/ the new encoding) so we don't have to re-archive everything in the past