Heikki,

These use cases are for Calendar dogfood only which assumes only events. We will have to do a whole other set of performance scenarios for the dashboard. Based on the dogfooding calendar size survey we came up with a 2000 item repository with 1500 events and 500 tasks. Since people will be subscribing to free-busy calendars in 0.7, I think we should push that up to 2000 events and 500 tasks. I think the upper limit on collections is 10. We expect people to have between 5 and 10 collections.

I will have to get back to you on the recurring event stuff. I think we need to work with QA to put together an appropriate test data set.

Sheila

On Apr 27, 2006, at 2:22 PM, Heikki Toivonen wrote:

What kind of repository should we measure against? How many tasks,
events, emails, etc? How many recurring, and how many of them in the
current day/week, how many in future/past? How large notes, attachments? What format notes, attachments, emails? How many collections? And so on...

See below how we stand with tests currently.

Sheila Mooney wrote:
+ Publishing and subscribing to shares

We don't have a performance test for this yet, but we do have a
functional test. Should be reasonably easy to do.

+ Synching calendars - can't do anything else while synching (will be
partly addressed with background sync)

I don't think we have a test for this either.

+ Switching between calendars

Seems we don't have a performance test for this yet. The closest we have
is a test to switch to All view.

+ Switching between app areas

Like I said above, we have a test switching to All view. Would be
straight forward to implement other cases.

+ Activating/Deactivating calendars - clicking on the checkbox
    - Since the checkboxes are slow to check and uncheck, people keep
clicking thinking it doesn't work.

You mean overlaying calendars? If so, we do have a test for this.

+ Importing a calendar takes a while - I can't do anything else with the
app

We have a test.

+ Double clicking to create a new event on the calendar
    - This takes some time to respond and people often do it twice
(thinking it didn't work the first time) so then you get 2 events and
have to delete one.

We have a test.

+ Drag an event to the Trash - it doesn't disappear immediately.
+ Dragging on the calendar canvas to resize an event
+ Dragging events on the calendar to a different - day/time.
- I find it easier to edit the detail view because I always end up
putting it in the wrong place.

I don't think our testing framework allows us to simulate dragging at
the moment.

+ Switching between day and week view.
- I clicked on monday (to display day), then Tues, then Wed. I had
to wait for it to display all of them.

I don't think we have a test for this.

+ Changing the calendar timezone isn't bad - could be some visual
feedback that it's "working".

No test yet.

+ Changing the timezone for an individual event (or making it floating).
This for some reason seems slower than changing the entire view.

No test yet.

+ Clicking the all-day checkbox to make a regular event and all- day one
(and vis versa).
+ Any edits in the detail view and having it update on the main calendar
    - Changing date/time
    - Changing the status
    - Changing the title

I think we have a functional test that does at least part of this, but
no perf test yet.

+ Moving back and forth week to week using the arrows at the top of the
calendar view.

I think we have a perf test to jump forward one week.

+ Clicking on the Mini cal to navigate weeks in the calendar

No test for this yet.

+ Stamping and unstamping events as a communication - to send event
notification.

Not sure what you mean.

+ Stamping and unstamping events as tasks.

We do have a perf test where an item is stamped, but I don't think we
have unstamping performance test.

+ Drag and drop events off the calendar canvas
- So slow that it's hard to know it's working. People often have to
do it a couple of times.

Like I said above, I don't think our framework allows us to similate
dragging.

+ Selecting an event on the calendar

I think we do this as part of some other tests.

+ Deleting an event
    - using delete key or menu items

We have a test but disabled due to some bug.

+ Launching and quitting the app

We have launch perf tests, but no quitting tests.

--
  Heikki Toivonen


_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Open Source Applications Foundation "Design" mailing list
http://lists.osafoundation.org/mailman/listinfo/design

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Open Source Applications Foundation "Design" mailing list
http://lists.osafoundation.org/mailman/listinfo/design

Reply via email to