[Epiphany] More Topic/Keyword/Folder madness! Oh, no!

[Jens Knutson <jensknutson yahoo com>]

(Sorry this is so long, but I've put a LOT of thought into this, and I
really think this is about as succinct as I can make this without
removing important information!)

Ok, after watching some of the bookmarks discussion here, and reading
everyone's email, and checking out all the links, I have to say I'm
really not sold on the keywords idea.  If this email seems biased, well,
it may well be so. :-)  I just think the keywords system is a mistake,
and I want to see Ephy succeed, like it deserves to!

The supposed advantages of keywords seem to be as follows:

1) Avoids hierarchy, which is confusing for some (many?) users, ie: it's
a simpler model
2) Provides incentive for users to organize bookmarks through this
simpler filing model
3) Somehow superior search functionality?
4) Ability to have several keywords assigned to one bookmark, enhancing
the browsing of bookmarks

Of all the arguments I've seen, these are the only ones that seem unique
to the model itself, as opposed to the *implementation* of the model. 
All of the other arguments in favor of keywords are things that I see as
implementation issues.

I'd like to address these advantages now.

The first advantage, that of avoiding hierarchy through a flat list
model, and thus gaining simplicity, is certainly admirable.  I'd
definitely agree that an imposed hierarchical model can be a strong
burden on end users, however, I fail to see why avoiding it is an
inherent property of the keyword model.  The folder model is
hierarchy-*optional*.  No one is forced into creating a hierarchy - the
user would have to do this themselves!  Thus, I believe the first
advantage of keywords to be false - it is not an advantage of the model,
it's an advantage of implementation.

The second keyword advantage of providing incentive to organize through
a simpler filing model, is flawed as well.  I don't think we're any more
likely to persuade users to file their bookmarks with keywords any more
than we are with folders.  Again, I believe this to be an implementation
issue, not an inherent property to the model. 

Thirdly, there's nothing about keywords that makes them inherently more
searchable than folders.  In fact, I think it might be interesting to
pursue a hybrid model, whereby a user can put a bookmark anywhere in a
folder hierarchy, and at the same time, assign keywords to a particular
bookmark.  Of course, this may be too complex, and prove to be the
Achilles Heel of my whole argument, but we'll see, eh? :-)

The final advantage to keywords is the ability to have several keywords
assigned to one bookmark, which enhances the browsing of bookmarks. 
This is nice, but I have a feeling that there is going to be a very
small intersection between the groups of users that assign more than one
(if any!) keyword to a particular bookmark, and the group of users that
don't understand hierarchies.  Also, if one does assign multiple
keywords to several bookmarks, one's keyword list can grow enormous very
quickly, thus effectively slowing down the browsing of bookmarks quite a
bit.  (See my screenshot here of how long a keywords list can get in
only a few weeks - http://upevil.net/ephy-keyword-list.png - note the
length of the scrollbar.)

So you can see why, from my perspective, keywords offer little advantage
over folders.  However, I believe that the problem with keywords is not
that it isn't good enough on its own, but that we in fact *lose* a lot
of important things by moving to such a model!  Here's a list of the
stuff we lose by moving to a keywords model - note that these things are
inherent to the model - these are not implementation problems, IMHO.

1) Lack of familiarity.  Every single browser I've ever seen or used
uses folders.  Most notably, evry version of Netscape, Safari, Opera,
Galeon, Konqueror, and the big behemoth, Internet
Exploder^H^H^H^H^H^H^H^H... er... Explorer, which commands a HUGE
percentage of the current userbase of the web.  While we certainly
shouldn't just ape the features and model of any one browser, IE in
particular, it is important, to quote Havoc, that we "spend our
different points wisely."  If we're really going to break from the
folders model, we'd better be damn sure it's worth the lack of
familiarity for anyone coming from another system, which *will* have
used a folder system.

2) Loss of the ability to choose a hierarchy.  The folders model can act
much like the keywords model, if we implement correctly, but the keyword
model is inherently lacking the ability to do "sub-keywords" - instead
of having optional hierarchy, we have no choice but to use a flat
structure, which can be really cumbersome as one's bookmark list grows,
especially if one uses more than one keyword per bookmark!  Again, see
the screenshot referenced before.

There are some other issues I worry about with keywords, like fast
access - using the current bookmarks window takes FOREVER compared to
using something like a folder system in a bookmarks menu.  I realize
that my real complaint here is "i want to be able to access bookmarks
quickly", but I can't think of any implementation of keywords that could
possibly work as well to access bookmarks quickly, using only the mouse.

So what do I want?  Well, in summary, yes, hierarchy can suck, but I
think it certainly Sucks Less[1] than the limitations imposed by
keywords, especially when keywords can be integrated into a folder
system, and because hierarchy is always optional.  After thinking about
this for days, and reading over everyone's materials, and trying out
really nice folder implementations, like Phoenix's and Safari's, that we
should shoot for a really smart, bad-ass folder implementation. 
Keywords are simply too restrictive and too untested.

That's it!  I'm done!  Please, everyone, let me know what you think.  Am
I a subjective, cranky, stubborn holdover for folders, or do I really
have a point here? :-)

 - jck


[1] This principle (well, this exact phrasing of the principle) comes
from JWZ.  Specifically, this page - http://www.jwz.org/doc/linux.html. 
The relevant bits are these lines, where he talks about how Unix (and,
thus, Linux) suck:  "Yes, that's right, though I've been living in Unix
for more than a decade, I think Unix sucks. Read the ``Unix Hater's
Handbook'' if you want to know why. But I'd rather run Unix than Windows
or MacOS any day, because Unix sucks less. That doesn't mean it doesn't
suck."

-- 
"The reasonable man adapts himself to the world; the unreasonable man
persists in trying to adapt the world to himself. Therefore, all
progress depends on the unreasonable man."

 - George Bernard Shaw