davical/testing

Regression Testing
==================

At present these regression tests are basically written to work in my
own environment.  While I am, of course, happy to see patches that make
them more generic they are still very much a work in progress.

Mulberry
========
At present the most demanding client to support is Mulberry, so the first
set of regression tests imitate Mulberry taking RSCDS through it's paces:

  1. Initial OPTIONS request at the root
  2. Initial PROPFIND request at the root with Depth 1
  3. Second PROPFIND request at the second level
  4. MKCALENDAR request to create a calendar at /user1/home/
  5. Third PROPFIND request duplicating the Second one (but finding a calendar now).
  6. Fourth PROPFIND request solely looking for the new calendar, requesting 'getetag'
  7. Not that Mulberry would let us do this, but we try to MKCALENDAR again at /user1/home/ to check for the error we expect.
 10. PUT our first event into the calendar.
 11. PUT the same event a second time, which should not give an error, but should respond with 'Replaced' rather than 'Created'.
 12. PUT a second event into the calendar.
 13. PROPFIND which should now show us both events.
 14. PUT an update to that second event.
 15. PROPFIND which should show us both events, with changes.
 16. MKCALENDAR somewhere else that we have rights to do so.
 17. MKCALENDAR somewhere else where we do not have rights, and which should fail.
 18. PUT into the other calendar we have just created.
 19. OPTIONS request against an illegal path, which should fail.
 20. DELETE the first appointment we created, but with an If-Match header that will cause it to return a 412 Precondition Failed.
 21. DELETE the first appointment, but this time with a correct If-Match header, it should succeed.


Evolution
=========
Evolution does things quite differently to Mulberry, only doing one OPTIONS request to confirm that
calendaring support is available, and then going for REPORT and GET.  It maintains an internal cache
so does not GET events listed in a REPORT unless they are new or changed.

 100. Initial unauthenticated OPTIONS request.
 101. Initial authenticated OPTIONS request.
 102. Initial REPORT request.
 103. GET the second event that Mulberry put there earlier
 104. PUT an event in the way style Evolution uses.
 105. REPORT which should show both the Mulberry and the Evolution events.
 106. GET the Evolution event we have added.


Mozilla Calendar
================
Similar to Evolution, Mozilla Calendar primarily only does OPTIONS/REPORT/GET/PUT/DELETE however
it does not have a cache, so its REPORT requests (a) request data to be included and (b) apply
a date range in their response.

 200. Initial unauthenticated OPTIONS request.
 201. Initial authenticated OPTIONS request.
 202. Initial unauthenticated REPORT request.
 203. Initial authenticated REPORT request for events from 9th October to 9th December which should find one Mulberry and one Evolution event.
 204. REPORT request against the second calendar created by Mulberry which should only find one Mulberry event.
 205. PROPFIND request which only 0.4 and later versions of Mozilla Calendar will do (see bug 355270).
 206. PUT a recurring event.
 207. REPORT on a period which will only include a subsequent instance of the recurring event.
 208. REPORT on a period after the end of the recurring event, which will return an empty result.


Chandler
========
Support for Chandler is still under development.  It appears to operate somewhat similarly to
Mulberry, although apparently without support for MKCALENDAR at this stage, and it also tries
to use more basic DAV functionality than other clients.  Basic operation appears to be OK,
although it appears to write some proprietary information into a ".chandler" sub-collection.

 300. Initial unauthenticated OPTIONS request.
 301. Initial unauthenticated HEAD request.
 302. Initial authenticated OPTIONS request.
 303. Initial PROPFIND request.
 304. Subsequent PROPFIND request which endeavours to retrieve permissions.
 305. Another OPTIONS request looking at the ".chandler" subcollection.

--->> To Do
 Lots of things, including GET, PUT, DELETE and encouraging Chandler to create the .chandler
 collection in order to see how it does that.  At the end of these regression tests so far,
 'Chandler' has still not discovered what events exist on the server, but in reality it does
 successfully do so.


Your Favourite Client Here
==========================
I would like to have more client software available to test RSCDS against, but so far it's
just these ones.  If you want to point me at more free software, or send me free copies of
proprietary software, then I will add it to the list as well as make RSCDS work with it.