Brion Vibber schreef:
When that happens, it's not an attack on the committer, nor on the submitter of the patch. No malice is intended; no punishment is meted out.
I didn't take it that way.
The basic problem with the hook as written was that it passed in only one of the many relevant bits of a title -- the dbkey portion. There's also a namespace, an interwiki prefix, and a fragment; they would only be manipulable to the hook by poking into internal members of the title object, which isn't a clean interface and could lead to compatibility problems in the future.
Those local variables are scattered all over the function. How about having a hook for each of them? (A hook to interpret the NS, one to interpret the interwiki prefix, one to generate the DB key, etc.) That way we could still keep the hook you reverted and rename it to TitleSecureAndSplitDbKey.
Roan Kattouw (Catrope)