Jay Smith wrote:
> On 03/10/2010 03:27 PM, yahvuu wrote:
>> Martin Nordholts wrote:
>>> 2010/3/10 yahvuu <yah...@gmail.com>:
>>>> Each of these dialog options points at a potential interaction problem. If 
>>>> the
>>>> dialog remembers an option, the user also has to remember that option.
>>>> In general, this amounts to additional cognitive burden to keep the mental 
>>>> model
>>>> in sync with the application state.
>>> Hmm I don't understand, how would there be additional cognitive burden
>>> on users if a dialog remembers a setting across invocations?
>> You are right, that was an invalid generalisation. I was thinking of the 
>> 'New Layer' dialog
>> where the 'fill' option tends to get in the way.
>> (If i leave that option to a fixed value -- like a prefs item -- there's no 
>> problem, i can
>> just hit  enter to create the new layer. If i, however, do change this 
>> value, i'd better
>> remember this the next time i create a new layer.)

> I am not convinced that the 'New Layer' example given presents a
> "better" situation as described.

well, it additionally depends on the task at hand -- which tells that it was 
bad idea
to hold the dialog state responsible for getting in the way. This cannot be 

Allow me to rephrase my original point:
Each of said dialog options can potentially get in the way
of smooth workflows and is worth being questioned.

Now for my example where it's indeed the dialog state that is 
For the New Layer dialog, the crucial question is wether the next layer will
be created with the same fill as the last one. If that assumption holds true
most of the time, then a stateful option is useful, as you described here:

> In the 'New Layer' example given, if I am doing a certain repetitive
> task, it is *highly* likely that I will want the new layer to have the
> same fill every time I do that function.  There is a significant
> "burden" in having to change this setting _every_ time.  (And to make it
> worse, some of these settings are not so easily accessible via keyboard,
> thus wrecking my shoulder from mousing too much.)

If however, say, a mixed stack of 10 transparent layers and 5 colored layers is
to be created, then using the stateful option becomes a burden: The user is 
to read and potentially adjust its value for each new layer.

In this case, it is more efficient to leave to the fill option to 'transparent'
and manually fill the layer later on, if required. Why? Because it can be done
blindly, at least when using the keyboard (see below).

> So, which is better:
> a) Knowing that the program will always force a default value and having
> to change it much of the time (in my case for Canvas Resizing, ALL the
> time).
> b) Knowing that the user is responsible to paying attention to what the
> value says when they get to the dialog and if it is correct for the task
> (which it will then be, once the user has set it, until later changed by
> the user).

Sorry, i cannot give a useful general answer here. Consider e.g. the
Gaussian Blur filter, whose radius setting matches both a) and b).

For the New Layer dialog, i'd prefer:

c) Always create a transparent layer, without showing a dialog. Then let the
   user fill the new layer if desired (not considering layer size here).

   This works nicely for a keyboard workflow: CTRL-SHIFT-N, CTRL-.
   Fast, no confirmation required and no application state has to be remembered.
   An open question is how to make this fast for a mouse only / tablet user.

For your original case of Canvas Resizing, i fully agree that the dialog should
remember the 'resize layers' option (until we get auto-sizing layers, of 


Gimp-user mailing list

Reply via email to