Bug#307768: exim4: check_local_user should reject system users?

[Marc Sherman]

Marc Haber wrote:
> 
> But the real-* business has been in the exim packages for a _very_
> long time (this is even part of exim3 on woody) and probably nobody
> knows any more why we include this. Does upstream recommend this in
> its default config, or is this a Debianism?

The use of real-* comes straight out of the upstream docs for 
skip_syntax_errors:
http://exim.org/exim-html-4.50/doc/html/spec_22.html#IX1867

In those docs, it does recommend that the real-* router be put 
immediately above the userforward router.

While the real-* router in the docs doesn't do any of the local-sender 
only verification that I propose, there is a strong implication there 
that the only reason to have the real-* router is to handle syntax 
errors in userforward, not for arbitrary third-party usage.  If 
anything, I'd suggest that

	condition = ${if def:$sender_host_address{false}{true}}

be added to the router in both the debian config and the upstream docs.

> If it's a Debianism, then we should probably think about doing the
> change that you propose. May I ask you to present your case on
> pkg-exim4-users or pkg-exim4-devel to get a broader opinion?

I'm not subscribed to those lists, but you're welcome to repost this 
thread (or just a reference to 307768, which should have all the 
discussion) there if you like.

- Marc