Re: Browser re-org

[Bryan Clark <bclark redhat com>]

Update.

So I've been working on this off and on for a little while.

So far I have a plugin for storing page information into an sqlite
database which does page indexing of history and allows for bookmarking
into this database.

Here's the scenario that I'm designing towards right now.  This is kind
of an ideal initial scenario that should lead up to others later.

Search Scenario

	New Window Opens

	Person enters a search
		- We save search terms

	Person tries a few results, no success
		- We save results as history objects

	Person refines search
		- We save refined terms with higher ranking than first terms

	Person finds a link of interest and saves it
		- We save this history item as a bookmark (aka highest ranking)

We now have encapsulated this entire scenario into a web of data objects
that can be searched for and each item saved can be found related to
each other.

Once this part is finished up I need to work on the SQL query for
searching terms and how to display relevant similar results.  I've built
a little test browser application using PyGTK and gtkmozembed and is
hooked into the sqlite database so I can test this search result stuff.

The main interaction for the browser is to have a single address /
search entry area where you type in a web address or search term.  If
the person enters a search term you send them to google, if they enter a
web site you feed that directly to mozilla.  I see the problem of
identifying what's a search term vs. what's a URL and it does seem to be
a difficult one.  However I think I've gotten around it with the dynamic
searching and analyzing the entry against your history and bookmarks.
I've yet to test this theory out, but I'm hoping it will work. ;-)

To explain that a little more, I'm storing all the bookmarks and history
in the same database (as pretty much the same objects).  When you type
"smeagol" into the entry area subsiquent matching resuls are shown in
the browser window as you type each letter; this is dynamic querying.
If you see what you want listed you'd hit "tab" or something to jump to
that and press "enter" to go to it.  If there were no matching entries
in the history or bookmarks you'd know by the time you hit enter since
no results would be displayed and you'd just have a "Search Google for:
smeagol" showing.  Similar to the keyword search / address entry right
now, but the feedback is done differently.

Part of the different feedback is the results view taking up the whole
browser window.  I'm assuming that when you start typing in the address
bar you are planning to lose the current browser window.  (understood
ctrl-enter does differently, this is still possible)  The results are
based off of the index of text in the site, not just the title.  I'm
hoping to get page thumbnailing stored with pages and displaying in
results sometime in the future, need caillons help for that one.

The goal of the search scenario is to enable people to type in similar
search terms to their original one and see pages they bookmarked related
to search to those terms before.  Plus pages they saw along the way to
finding that bookmark.

One other thing that I need some help figuring out how to do is to only
get the text of the part of the web page that people have seen.  This is
a tuning of the other idea towards this "Remember the stuff I've seen"
goal.  If I haven't seen it we might want to remember it, but not as
high a ranking as the stuff i have actually seen.

I'll probably have all this code online sometime soon in my baz
repository so others can play around.  Right now there's just a little
too much churn to share it and be useful.

Any thoughts?
~ Bryan