Re: Summary of Issues Raised to date

Laird Breyer wrote:
> There's one issue missing in your list, which came up in the header 
> compendium. POPFile also modifies the subject directly, which as I
> understand is necessary for filtering in Outlook Express. 
> There may be other filters who do this too.

In some configurations, SpamAssassin does this too. However, its spam
subject tag (in the configurations I've seen) is along the lines of
"*********SPAM*********". I've also seen it use "******OPT-IN******".

> I would think that this practice is a prime candidate for
> standardization. Here are some issues with it:
> 
> - Spoofed [category] tags. How does a filter deal with them?

If the filter is closer to the user in terms of control than the source of
the existing tags, I think it would be correct to remove existing tags and
add new ones as appropriate.

> - Population explosion of [category] tags. In an email conversation,
>   the tags become a nuisance just like Re: Re: Re: ...

Ideally, each user would only have one piece of software doing this tagging,
preferably after any subject-maintaining software does its work. Thus, no
email should get more than one of these tags unless the end-recipient's
software decides that it merits more than 1.

This comment brings to mind another issue:
How should tags in the subject following 'Re:' be treated? From the end
recipient's point of view (I guess the that would be conversation-initiator
in this case), it seems correct in msot cases to remove such tags. Such a
user doesn't care how his correspondent tags his messages. Also, in terms of
threading with outgoing messages, it's easier if the response subject lines
up with the originally transmitted subject.

> - If the tags are modified over an email conversation, or added late
>   say, some of the threading algorithms are affected. For example, 
>   Jamie Zawinski (http://www.jwz.org/doc/threading.html) describes a
>   robust way of threading, which AFAIK is widely used. However, the
>   algorithm sometimes compares subject strings directly to see if the 
>   messages belong to a thread. Inserting/removing [category] tags in subjects
>   could play havoc with the algorithm.

I note that looking at the subject is the last real step of that threading
algorithm. It should be a simple matter of programming (:-)) to make it
ignore any tags it sees, once there is a standard for their construction, or
correlate them, for that matter.

> - The tags can be skipped by the originating algorithm, but
>   independent filters have no way of recognizing them (ie the [,]
>   markers are much too common in mail subjects)

That's why we should come up with a standard for the construction of these tags.

> It's rather distressing now that I've made a list of some of the problems.

Making such lists is usually a worthwhile exercise. Even better, it seems
that most of the issues you presented have a tractable solution.

Philip Miller