Re: [Epiphany] Bookmarks

["Marcelo E. Magallon" <mmagallo debian org>]

 > A good try at proving why hierarchy isnt the best way is here:
 > http://www.gnome.org/~seth/storage/associative-interfaces.pdf Mostly
 > to show a good, rational method of dealing with usability issues.

 Absent from that draft (I wouldn't dare call this a paper) are:

    * Use Cases

        It's full of "we believe"s.  Fine, you believe that's the case,
        but please show some evidence to back the assertion up.

    * References

        Only three of the provided references are refereed papers, and
        the newest one on this batch is 12+ years old.  I don't consider
        a Master's Thesis a refereed paper, it wasn't read by enough
        independent experts on the field.

    * An affiliation

        Anyone can assert whatever they please.  But I want to know
        who's making the assertions.

 I have a problem with this in particular:

    It is not uncommon to encounter a cognitive justification of the
    hierarchical filesystem.  The argument goes that search operations
    require recall, because you are not presented with a list of choices
    to choose from you must remember attributes of the file, whereas
    "file browser" interfaces require recognition.  Because is a well
    established cognitive phenomenuum (sic)that people can recognize a
    greater number of items far faster than they can recall the same, it
    is presumed that people will be faster and will be able to handle a
    larger number of files with a file browser interface. [...] search
    is more accurately characterized as an associative task not a recall
    task.

 Fine.  But the author is also applying his conclusions to the wrong
 domain.  Searching unknown corpuses *is* an associative task, but here
 we are talking about searching a corpus that the user himself has
 created.  The contents are not completely unknown to the user.  "Where
 did I put that piece of information?" is the question at hand.

    At the other extreme, if you allow people to denote an object by any
    attribute they find relevant people are extremely efficient (more so
    than even recognition). This is because in order to have a goal of
    "open file" in mind the person has some characterization of the file
    in their head.

 but then you are shifting the problem to a different level, not solving
 it.  Now the key assignment becomes the problem.  I've seen it with
 Epiphany: assigning topics to the bookmarks becomes tedious after a few
 dozen bookmarks because the list of topics grows with time, and the
 topics you came up with a the beginning aren't as good as you thought
 them to be.  If the contents is magically characterized (which seems to
 be Storage's goal at some level), this works, but who's going to come
 up with the spells?

 The author does present some nice arguments, but I'd wait for this work
 to be reviewed by a group of experts in the field before using it to
 backup design decisions, and I'd particularly avoid this as a
 justification for inflicting Epiphany's bookmark system on users.

 Marcelo