Re: [Evolution-hackers] Category Searching

[JP Rosevear <jpr ximian com>]

On Tue, 2004-02-03 at 02:19, Bryan Clark wrote:
> On Mon, 2004-02-02 at 12:10 +0200, Tuomas Kuosmanen wrote:
> > Hm, what if we include all categories the user has "seen" - I mean, all
> > of the "native" evo-supplied ones plus any category that has been seen
> > in an address card in addition to those? So that Evo would "remember"
> > any special ones once it has been encountered once..? Is this something
> > that could work?

The main problem here would be LDAP currently, the addressbook does not
by default load all cards for LDAP.  This means no categories could show
up until you searched for LDAP entries in some way other than by
category (ie by last name, at which point any categories on the returned
cards would become "known").

> I like this idea, it would keep the list simple until the user
> complicates it on their own with lots of their categories.  Somethings
> related that might need to be worked out would be: How is the category
> list manageable?   How long is a category kept after it hasn't been
> seen?  Do we keep a category that's only been seen once?

I like this concept, but I'm not convinced on the practicalities of it.

> But I think the idea is in a good direction for this problem.
> 
> > The long list issue is something that sucks though, do we need to use
> > autocompleting combobox-like widget instead of the option menu?
> > Something like the composer uses for To: -field etc?
> 
> I disagree with this, Category systems are not really triggered via
> recall methods, rather recognition (i.e. one does not remember/'recall'
> all the categories they have created, but 'recognizes' them when they
> see them and probably recalls what they are for after).   
> 
> The composer To: field can rely on user recall since people need to know
> who they are emailing almost before they start and as long as they have
> a cue that is close to the address, like name/nickname/email-address the
> auto-complete can assist with getting the address exactly right.
> Categories on the other hand are not as easy to remember as peoples
> names, for obvious reasons like people are real things that you might
> have some kind of relationship with vs. categories are made up meta-
> information types.  Imagine if it was Work - Project A or Ximian -
> Project A... each category require you to type ahead... then delete and
> try again, maybe it was Novell - Project A? :-)
> 
> I'd recommend a Category system using drop-down options.  In order to
> lighten the amount of categories I think we could do a few things.  
> 
> 	1. Go with a small finite set 
> 	2. Create easy way to modify the set
> 	3. Use the 'seen' method tigert suggested
> 
> Maybe a combination of those?  1 + 3?  There's probably lots of other
> ways, if we worked on some use cases for this it could simplify things a
> lot.

Good points.  However I'm thinking that the small the finite set the
more categories people will add, so perhaps the list will grow too long
anyhow.

-JP
-- 
JP Rosevear <jpr ximian com>
Ximian, Inc.