Browser re-org

[Bryan Clark <bclark redhat com>]

So I've been slowly interviewing a number of people about their web site bookmark habits.  I'd like to share a little of what I've been talking to people about, what I've found and what I think could be an interesting thing to start on.

The process

So I probably talked to about 20 people total.  Random people I've met, family members, co-workers, and others.  The backgrounds on all these people vary from hard core programmer to grandmother and military sergeant. 

I started by asking about how many bookmarks people have.  Then depending on where that took I would often ask about the bookmarks they use the most.  If they organize their bookmarks at all, if so, how do they do that.

My findings

>     Many started out with few bookmarks and then gradually hit a point where they had so many that the bookmark drop down list was huge.  
>     
>     Most people didn't categorize their bookmarks at all, if they did do it, then they categorized them at the time they found them.  There was one person who would open up the bookmark editor and categorize them all every so often, usually when things looked cluttered.
>     
>     Of the categorizers, many said that after a while they couldn't remember what the original category of a bookmark was.
>     
>     Of all those people, even the meticulous organizer said that he couldn't always find the bookmarks he was looking for when he wanted them.
>     
>     About less than half used the bookmark bar, although they didn't know it was called that, to keep often used bookmarks handy.  However even that got cluttered over time for many of those people.
>     
>     Most of the power user types said that they had given up on bookmarks completely and only use Google to find stuff instead.  They claimed it was fast and required zero organization on their part, but usually allowed them to continually find their target.
>     
>     Many non-power user types said they used and liked the IE bookmark sidebar, because it was an easy way to get to all of their bookmarks.
Time to cut out middle management (this is the re-org part)

So, from all of this I've started to conclude that bookmarks as we, and firefox use them are becoming a burden on people and no longer useful.  Google is doing some interesting things with search history now, where they record what links you click on after searching.  It provides a pretty good layer of links that could have been important.

Part of this is why I've been against letting any bookmarks system get into Evince, I do not want to create _another_ thing that people have to manage on their computer.  Another set of bookmarks is the last thing people need, and I don't think that combining all bookmarks into a single system is necessarily the right solution either.  Maybe later... but not yet at least.

So I think we should look at doing something to change the way bookmarks are used.  I haven't figured out a UI for this yet, but I think that's something that we can create, tweak and iteratively build after we understand what we're doing.

Bookmarks and History, a match made in heaven

So my idea about bookmarks is that we need to create a way to use more contextual information in allowing people to retrieve them.  When I was looking for the pygtk tutorial that showed good examples of creating a web browser with glade + pygtk I didn't really want to search google for that.  Google would give me all the possibilities in the world (wide web) of those things, and not the ones that I've seen.  The page I had seen once is the one that I wanted to get back to.  When I'm looking for new things, I use google, but it's just because I ca'nt find what I want in bookmarks right now that I have to use Google to look for things I know I've seen.

When I originally found that page, I think I had searched google for pygtk + mozilla and wandered around the web until I found that one.  Which I believe I bookmarked.  Of course my bookmark didn't take into account any of the information leading up to my bookmarking it.  The search terms I used to find it, how many other pages I looked at along the way, and lots of other 'meta' information was lost.

I'm proposing we keep all that information.  We look at the web and our browser as a couple different use cases and really tune for those things.  Off the top of my head I use the browser for:
>     1. Checking pages I look at often: Google News, Planet GNOME, Evince Bug Listing
>     2. Finding new things through google (insert other search engines here) to solve a problem I'm having.
>     3. Opening links someone sent me: link pasted into IRC chat
>     4. Going back to things I found once before: bookmarks for code reference pages
If we look at these 4 generic uses (there's probably a few more) and tune our bookmarks system for this, I think we can provide a much better experience for several aspects of the browser.  

Crazy Research Thoughts:

For 1. maybe we could automatically put those items into the bookmarks bar for people
For 2. we remember the search terms used and the pages clicked on in the search and store/display that so people can more easily search again for those pages
For 3. maybe we can know where the browser was launched from and save that so that we can store "link was found from #epiphany had this link 3 days ago"
For 4. Maybe this is where we create a search applet that searches history + bookmarks and shows results in a dynamic query system or forwards the search on to google if nothing is found locally.

What I'd like to do is (maybe) 2 things.  

First start working new ideas for handling bookmarks, I think there's lots of ways to look at bookmarks and page history data, like I mentioned.  But we need to work out the ideas and then design based off that.

I think it's very important to drop existing assumptions about what a browser is in this project.  This is why my second point is to create a little experimental project that allows people to drop those assumptions and work on new ideas.  It's hard, if not impossible to be creative when you have all the constraints of the existing system on you.  What's not intended with this is that we'll create something entirely new and scrap everything we've already done.  The idea is to branch, work out interesting ideas people have, then look into folding those ideas back into the mainline.

So the second piece is to start creating an experimental project, maybe we can base it off pygtk + pymoz to quickly work on experimenting with the browser.  We call this thing the epiphany-research module.  The whole purpose of this module is to have a testing and proving ground for new ideas.  I don't know that this module needs to reside in CVS, but we should have a common place for people to work on new ideas.

I've been trying to use epiphany-extensions for this, however it's not working out.  I can create menu items and a few other things that might be needed, but I can't do other totally crack things like replace the mozembed with a list view for search results.  And I can't seem to take over the URL entry, suffice it to say that there's lots of stuff I'm touching that I don't think epiphany-extensions was meant or wants to touch.

And finally

I'd be really cool to see lots of activity on this list related to these kinds of research ideas.  With a quick base to work off of I'm hoping that more people will be able to find developers to work with ideas and more people will be able to work out ideas of their own.

Perhaps it would be a good idea to pre-pend [RESEARCH] to your email subject if you want to discuss a research idea you're working on.  And even better, if you don't have a programmer to work with, pre-pend [RESEARCH-IDEA] so we know you're just proposing an idea that someone could work on.

Cheers,
~ Bryan