Re: bookmarks backend needs to be choosier

[Nat Friedman <nat ximian com>]

On Mon, 2003-07-21 at 05:17, Edd Dumbill wrote:
> So, an email I received has the character sequence '#9' in it.  The
> textchainer sent this as a bugzilla clue.  Then bookmarks picked out
> every single bookmark of mine with the character '9' in it.
> 
> Bookmarks should probably not listen to bugzilla clues.

Yeah, the bookmarks backend should work like this:

        * Match similar URL clues.  For example, if the user is browsing
        http://nat.org/dashboard/ and you have
        http://nat.org/dashboard/status.php3 bookmarked, that seems like
        a decent match.
        
        * Match text/plain and text/html clues against bookmark title
        and indexed bookmark content, using the indexer.
        
I think we should leave it up to the indexer not to match clues that are
too short, for example, or, rather, to downscale their relevance.

I've basically been waiting for the indexer to index web pages to fix up
the bookmarks backend.  Any other quick fix is likely to be wrong (e.g.
if you try to limit it to clue text >3 characters long, you miss the
"Xr" match that I screenshotted a while ago).

Part of the fix here might involve the ContentName/ContentType
distinction I mentioned last night.

Nat