Things to consider for future releases
Clean up Remote Authentication
With the addition of OpenID and LDAP support in Geeklog 1.5.0, remote users will become more common and existing problems will become more apparent. For example:
- Duplicate email addresses
- Missing email addresses
- Ability to make remote users "local" users (and vice versa)
- Need to clean up and unify / merge the login forms (side block and users.php)
- Add a "service" column in the Admin's user list when remote auth is enabled to easily identify remote users and the service they're using
- Ability to re-sync data with remote auth service, e.g. change of email address
- Clean up user preferences ("My Account") for remote users: Hide options that are not available (e.g. password change)
- Also: Support for OpenID 2.0 would be nice (Geeklog 1.5.0 supports OpenID 1.1)
The configuration GUI, introduced in Geeklog 1.5.0, will probably require adjustments once it's in widespread use, e.g.
- Re-ordering / re-grouping of options
- Ability to allow config access for certain user groups (currently requires Root access for everything)
It's about time we replace the md5 password hashes with something more modern and robust.
- Maybe use phpass
- To solve: What do we do with existing accounts?
Somewhat related: If we wanted to support WSSE for the Webservices, we would either need to store the user's password in a way that's secure but reversible (so that we can get the unencrypted password for use in WSSE authentication) or think of a way to create a token / temp. password that the user can use in WSSE authentication.
HTTP Digest Authentication
Alternatively (instead of WSSE), if we wanted to support HTTP Digest Authentication, we would only have to keep track of the (user:realm:password) hash, probably as a separate field in the gl_users table, that needs to be updated whenever one of the three elements changes.
Since kses is no longer maintained, we need a replacement.
- ??? (tbd)
Make better use of HTTP
Sounds odd, but there's a lot of room in Geeklog to make better use of specific features of HTTP, especially HTTP status codes and additional HTTP headers.
- HTTP status code 201: The proper response after creating something would be a 201 status code, accompanied by a Location: header pointing to the newly created resource.
- HTTP status code 303: When something has more than one URI, we should point the client to the preferred URI, using a 303 and a Location: header.
Example: when URL rewriting is enabled and the request was for the "non-rewritten" URL.
- HTTP status codes 4xx and 5xx: In case of an error, we usually only display a human-readable error message, but send a 200 status code. There's a whole range of HTTP status codes that would make more sense to use, for example:
- 400 Bad Request (invalid / incomplete data)
- 403 Access denied (obvious; already used in some places)
- 409 Conflict (e.g. when something already exists)
- 500 Internal Server Errror (sounds like we should send that whenever we display "Unfortunately an error occured ...")
- Make use of Last-Modified: header (where is this actually possible?)
- Output compression