I wrote (about graphical mail clients):
> None of them have proper support for mail scoring.[...]
> Mail scoring (based on sender, subject, spam scores, etc.) takes a
> little bit of setting up, but can be very useful on something like the
> Fedora lists if your client has it.
Les Mikesell replied:
> Where possible, I think this is better done at the transport layer with
> MimeDefang, adding headers that the user agent(s) can easily parse for
> rule processing.
This is true, and I actually do this to an extent. But it has problems.
For one thing, the transport layer doesn't have as much information
available -- it's nice to be able to score based on other e-mails in the
thread. For example, if the thread was started by Eric Raymond, then the
thread can be marked down as not that interesting (and sorted to the
bottom of the thread list). Then if "[OT]" starts appearing in the
thread subject, the chances of anything relevant being said are probably
low enough that it's not even worth reading the e-mail. So you need
support for "who started the thread" and "what's the subject line" and
the MTA Doesn't Do threads.
Cameron Simpson wrote:
> Depends how nice the scoring rule syntax is. And it's fine unless yu want to
> change the scoring system; then you need to refilter your email.
Another good point -- if you've got scoring, you might as well use it
for kill-filing.
Les wrote:
> That way you don't have to wait for scanning to happen
> at the user level
Cameron replied:
> True, but sorting on a score doesn't matter _until_ the user is reading
> the email...
Usually (in Mutt), scoring takes place using information the mail client
has to know anyway (name, subject, etc) to display the e-mail.
If you really do want to scan the whole mail for scoring purposes, then
yes that is better done at mail delivery time. (For example, I scan list
mail to identify posts that are HTML-only, and put that information in
X-Label, which Mutt will use for scoring purposes).
Les wrote:
> and aren't tied to the quirks of any particular mail
> program.
In my experience, the quirks of the particular mail program are likely
to be there anyway, and will affect the way you can use scores,
especially if the mail program isn't designed for scoring. You don't get
much MUA-independence by moving scores to fixed headers.
James.
--
E-mail: james@ | [Bradford Cathedral] took 194 years to complete. A
aprilcottage.co.uk | construction period of nearly two centuries may seem
| ridiculous to us, but of course builders were a lot
| quicker in those days. -- "ISIHAC", BBC Radio 4
