Re: [Epiphany] Bookmarks

[Marco Pesenti Gritti <marco gnome org>]

On Thu, 2003-10-23 at 20:25, Jon A. Solworth wrote:
> On Fri, 2003-10-24 at 00:54, Marco Pesenti Gritti wrote:
> > > 	I agree that hierarchy is not the desired property per se, but
> > > 	one way (not necessarily the best) of implementing it.  The
> > > 	properties that one gets from hierarchy is:
> > > 		1. Clustering: related material is together
> > 
> > In my mind a thing is often related to multiple other things, not to one
> > alone. So this doesnt work that well for me.
> > 
> 
> But isn't this just clustering in multiple dimensions?

Sort of. But the multiple dimensions are hardly representable in a tree.
For example a bookmark can be related to multiple topics, to the place
from where I launched it (an email for example, or the "parent" site),
to the time I visited it, to the sites I visited that links it ...

> > > 		2. Scan a short list and click: find things
> > > 			quickly
> > 
> > Well it's more like, scan, click, scan, click, scan, click, hrm I
> > thought I put it here !. Not exactly quick.
> > 
> 
> Basically, the issue is that a tree gives a search cost
> which is logarithmic in the number of elements.  Yes, I
> know that it doesn't work as well when it comes to ad
> hoc collections.  The issues here seem how much stuff
> you have to scan, how easy it is to scan, and how easy
> it is to click.  (Sometimes I seem to be clicking again
> and again answering small variants on the same question).

While ease of scan and click can improved a bit, if you have an high
number of items spatial organization is not going to work well enough
(the search cost is too big).
Also importance of recall shouldnt be ignored here. You never scan the
whole list, it would be too slow, you have a basic image of where the
item you are looking for is, even before opening the menu. But again
human memory works well with a low number of items.

I think there are different needs when retrieving personal information:
1 Accessing very fastly information that you need very often. There
spatial organization is pretty efficient. You have a low number of items
and your memory can dominate them making scanning and phisical movements
fast and sort of automatic.
2 Accessing important information on a longer time basis (for example
finding a site that took a lot of time to find, but that at best you are
going to need 1-2 times again). Using memory there is hopeless (memory
as spatial memory of where you put the item). Using rigid categorization
is going to be very slow because you have not memory to help you
scanning and clicking, and failure prone because there is only a way, a
path, often a pretty artificial one, to get to an item: the time passed
and there is a good chance that your mind will follow a different path
now. Also if you are using the same space of 1, you are going to slow
down it too.
3 Building new information and associations between already existing
items. That is the sort of idly browsing Google Directory someone
described. Even in that case IHMO a tree is a pretty poor tool. We can
store enough information about the user to be able to show him related
alternatives contextually. But anyway I'd argue that most people have
skills or time to build a well organized hierarchy that can be
interesting to browse to find new information.

> > > 		3. I would think that a fully automatic system would be
> > > 		 sufficiently inaccurate that an overide would be
> > > 		 needed.
> > 
> > Epiphany bookmarks arent really fully automatic ...
> > 
> > > No, I wasn't advocating a choice of 12 different bookmark editors. 
> > > But an effective bookmark editor is critical for heavy users.  
> > 
> > The problem is the definition of what is effective and what is not ...
> > 
> > > If
> > > you've got some very experimental ideas to try to move forward,
> > > great.  But also with experimental ideas, some work and some don't.
> > 
> > Exactly as of existing solutions some work and some dont ;)
> > 
> 
> No, I'm not arguing the status quo here. But doing something
> experimental is going to be different than tried and true.
> Invite over a bunch of friends and cook from recipes or try
> to wildly improvise.  The latter is probably going to have worse
> average quality but it may make up for it in other dimensions:
> more fun, discover something new, etc.  If your going to do
> experimental stuff more things can and almost certainly will
> go wrong.  Of course, its worth it if the search for something
> better is what drives you.  Moreover, lets face it we don't
> need another browser, so a new browser should be different.

If I'm getting you correctly here, we dont disagree at all ...

> Architectecture is about usability; its about good engineering
> construction (robustness, etc.).

Good point. I was just misunderstanding your intentions.

>   If you want to really push
> bookmarks, maybe the become more than browser URLs.  Maybe you
> want visualization tools, archiving tools, dynamic load tools,
> and expermenting with different bookmark organizations.

I and other developers would agree on this I think.

Marco