[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kmail-devel
Subject:    Re: Why, oh why? -or- scoring
From:       Guillaume Laurent <glaurent () telegraph-road ! org>
Date:       2001-06-01 18:56:55
[Download RAW message or body]

On Friday 01 June 2001 21:59, Marc Mutz wrote:
> > Because scoring can only be done within a folder, as rules are on a
> > per-folder basis.
>
> <snip>
>
> You can extract the folder the message is in with KMMessage::parent().
> Also, I see in the configuration dialog for scoring, that there are
> rules that apply to all "groups" (see, you didn't even change
> "group"->"folder").

No I didn't because I didn't write the dialog. It's defined in libkdenetwork, 
which knode also links with. Mathias Waack (the author) has agreed that 
there's a need for customisation here, yes.

That rules can be applied to all groups doesn't mean scoring is not per 
folder. Please look at kmscoring in libkdenetwork.

> Hm, maybe that's my intention? Maybe I'm thinking that calculating
> scores can be equally well and better done using KMails filters?

Think again. It can't. As it is, filtering happens when mail is incorporated, 
or manually. Scoring is performed each time you open a folder. It's meant to 
be much more dynamic than filtering.

> > As I said, that doesn't make sense (rules are on a per-folder basis).
>
> It would make sense if you integrated this stuff better into KMail. The
> "different scores for different groups" thing is fine for a newsreader,
> yet for a mail client, it's at most secondary. Here's why:
>
> In mail clients, the "groups" are not fixed by someone outside. It's
> the user himself (or herself) that defines (via filter rules), what
> gets to be moved into which "group". It follows that the user has
> defined a set of rules that applies to all messages in any given
> folder, except inbox. Thus, he/she can very well use these same options
> to define scoring rules.

No, you're still missing what scoring is for. Scoring typically applies on 
"subject" and "from" headers. It's mostly used to highlight threads you're 
interested in, or messages from "important" people.

Suppose you're subscribed to several kde mailing-lists, with a folder for 
each, and you want to score up messages from foo@kde.org in all of these 
folders, but nowhere else. How are you going to express this as filter rules? 

> Now you will say: "But that's duplicated efford. Why should I add the
> same rules for scoring _and_ filtering?" And you are _right_. That's
> the precise reason why I want scoring to be done by filters.

No. As you said, filters are for putting messages into folders. Scoring is 
for sorting messages within their folders. Scoring is not meant to decide 
where to put a message.

> The overhead currently stems from the fact that - for search rules -
> you currently would have to create a scoringmanager and a scorable
> article for every message you are given.

Not as it is. You create a score manager once (when you open the folder), 
then score messages if there are any rules in the score manager.

> As I understand it, scoring is a method to calculate an "importance
> measure" on the space of messages.

Yes.

> It works just like filtering.

Not at all. I reckon there's some overlapping between the two features, but 
they are still way too different. I suggest you take a look at Gnus, from 
which this feature is inspired.

Please show a little less haste in criticizing something because it doesn't 
fit your idea of how it should be. This feature is a nice example of good 
design leading to code reuse (I praise Mathias Waack here, not me). You're 
about to blindly break it out of misunderstanding.

-- 
					Guillaume.
					http://www.telegraph-road.org
_______________________________________________
Kmail Developers mailing list
Kmail@master.kde.org
http://master.kde.org/mailman/listinfo/kmail

[prev in list] [next in list] [prev in thread] [next in thread]