Re: GUI questions

[John Darrington]

On Tue, Oct 02, 2007 at 06:00:55PM -0400, Jason Stover wrote:
     
     On Mon, Oct 01, 2007 at 10:10:29PM -0700, Ben Pfaff wrote:
     > I wonder what everyone thinks about designing a "good" GUI (at
     > least, one that we think is good) versus designing one that works
     > the same way.  Do you think that users will be confused by the
     > changes, or do you think that they will appreciate any
     > improvements that we make?  It is hard for me to guess.  It is a
     > little like OpenOffice.org vs MS Office: OO.o can improve the
     > interface all they like, but MS Office users won't appreciate it,
     > because they're used to MS Office.

When OO.o and Gnumeric started, they tried to follow the MS interface
very closely.  As these programs have become more popular, they've
gradually started to depart from the MS user interface, and
incorporate their own ideas.   Psppire is a new project, and until we
get more users I think we should follow the Chicago UI in most cases.
     
     I'm not sure what the answer to this is. The problem is that none of
     us are GUI users. Like Ben, the only GUI I normally use is my
     browser's. In a message from a long time ago, John called the GUI the
     "methadone" for getting a user used to the syntax.  The problem is
     that only programmers think this way.

As discussed in another thread, a "good" GUI is designed such that
it's impossible for a user to make a mistake.  Ipso facto, it is
impossible for anyone to learn anything using a GUI.  For that reason
alone, I would have absolutely no objection if schools and universities
banned the used of GUIs.

However, a GUI is a good way to promote PSPP.  A good analogy is an
advert in a glossy magazine; the advert provides absolutely no
information whatsoever, but gives the read a feeling that the product
must be a good one.

     
     I've been thinking about this GUI both as someone who wants to code
     part of it and someone who listens to frustrated, non-technical
     users. (I'm on a university-wide "technology committee.") Is it
     possible to have a somewhat loose philosophy about the GUI? When writing
     GUI code, maybe we should:
     
     1. Write it as if it may have to be changed in the future, if
     necessary. Given our own lack of interest in using a GUI, we don't
     know what users will love or hate, so we should expect to make
     unexpected changes. The GUI should be written to facilitate such
     changes. (This statement may be too broad, but it seems like a decent
     goal if we aren't certain about how a GUI should look.)
     
     2. Don't imitate a bad feature of another GUI inless it seems users
     must have it that way, or if the feature is so common that changing it
     would annoy a lot of users. 


I concur on both points.  My suggestions for a pspp gui design
guidlines document, if we have one would be somthing along the lines
of:

*   Design the gui to look like that of spss so that users' are
    familiar with it, unless there is good reason to depart from this.

*   Good reasons include lack of features (for instance, I did
    something completely different with the Find dialog, because I
    thought the spss one was pretty bad).

*   Follow the Gnome HIG where they don't conflict with the
    above (the HIG says not to use frames; spss uses them rather
    extensively, so I have too.  I think the HIG is overstates the
    disadvantage of them).

*   It's acceptable to depart from the Chicago design for dialogs
    which are more than two levels away from the main window.  Eg,
    In spss, Ben's, /Statistics treeview in DESCRIPTIVES is implemented
    by a couple of sub-dialogs with checkboxes peppered over them.
    I think Ben's design is better, and it's unlikely to upset users,
    because they'd have to traverse at least two states to access that
    functionality anyway. 


     In my experience, almost all uses of a
     program like PSPP are accounted for by a few features. Users familiar
     with another program expect those common features to behave a certain
     way, but don't care if the less common features get moved around a
     little. So there doesn't seem to be any reason to duplicate bad design
     for seldom-used features. In my experience, most users use only
     DESCRIPTIVES, FREQUENCIES, CROSSTABS and GLM; utilities like
     RECODE and COMPUTE.
     
GLM is certainly very commonly used.  I know of several users who
regularly use RELIABILITY (to calculate Cronbach's Alpha).

J'

PS.  As a general point, I think that the GUI framework we have in the
posix world is very poor. See
http://www.xs4all.nl/~evbergen/groklaw-gui-terminal.html for some
ideas how it could be done better.


-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.



 LocalWords:  Guidlines

signature.asc
Description: Digital signature