IBM To Linux Desktop Developers: 'Stop Copying Windows' (InformationWeek) [LWN.net]

IBM To Linux Desktop Developers: 'Stop Copying Windows' (InformationWeek)

Posted Aug 10, 2008 16:15 UTC (Sun) by josh_stern (guest, #4868)
Parent article: IBM To Linux Desktop Developers: 'Stop Copying Windows' (InformationWeek)


Without intending any disrespect to the various participants in this discussion, the whole
thread strikes me as incredibly primitive.   People are arguing about good and bad software
without using any concrete definitions, any objective criteria, or even any good examples of
what they are talking about and what the supposed problems are.   Against that context,
discussion is just an extended venting of subjective opinion.   We might as well argue about
who recorded "the best song".

Science and engineering of a subject both have to start somewhere.  It would be better to
start with some simple, quantifiable models of the relevant phenomena even if they are
inadequate, because then they can be precisely debated and improved.    For example, one can
make a list of, say, the 100 most important/common tasks a desktop user should be able to
accomplish, list for each task the minimum number of interface gestures they need to use to
accomplish this task, and the minimum number of propositions they need to learn in order to
know how to complete the task with the minimum number of gestures - now the functional
complexity of the interface is "measured" as the union of those propositions.   Do you think
that model is bad?  Fine, propose a better concrete model and then someone else would argue
for an improvement over that, etc.   At least that would be the beginning of engineering
instead of the endless and completely unproductive venting of raw opinion about "usability".

(Log in to post comments)

It'll never work

Posted Aug 10, 2008 19:37 UTC (Sun) by krishna (guest, #24080) [Link]

I take issue with your model :-)     
     
Graham Nelson, who developed Inform 7, described why the dev process     
wasn't open to everybody:     
      
<quote>Had Inform 7 been developed in open source, I am fairly sure it      
would now be an elaborated version of the superficial prototype, and that      
it would be much the poorer. </quote>     
     
Someone who puts out the first model will have people elaborating on the     
prototype, not necessarily refining or disrupting the model to produce a     
new one.  I think this is because developers don't have respect for what    
it takes to produce these models, and will complain about not being able    
to write code against a continuously changing model such as (dare I say    
it) the Linux Kernel ABI, to provide an even more concrete example.    
    
Frankly, I think there should be a universally accepted handful of books    
about UI design that developers point new developers towards; just 
validating the ideas of measuring number of clicks, time-to-task, facts 
about human visual and motor processing would be valuable.  These are 
something I'll bet very few developers get encouragement from the 
community as something valuable to learn.  
  
And my two cents about existing UIs; I really like what KDE has done in  
centralizing keyboard shorcuts, mime-typing, etc into the control panel   
in such a way that they can be accessed across all applications, probably  
even non-kde ones.  With KDE 3, they provided a slider bar to match level  
of UI effects to the power of your computer; AFAIK, to do this in Vista  
you need to buy a whole different version of the software or start hacking  
the registry.  So in that regard, KDE itself is a set of good models that  
can be complained about or improved.  So yeah, it could work :-)

It'll never work

Posted Aug 11, 2008 1:46 UTC (Mon) by josh_stern (guest, #4868) [Link]


I think you misunderstood the sense of model I was using.  I wasn't talking about "model code"
- e.g. a prototype.   I was talking about a theory of what it means for one interface to have
better usability than another.  Instead of Person A saying "I like GUI1 better" and Person B
saying "I like GUI2 better" we should have Engineer E saying "I'm going to use Model 1 to
define and measure the usability of GUI1 and GUI2 in order to quantify which one is better"
and then Engineer F can come along and say, "No we should use Model 2 to do that".