[Mailman-Users] sendmail bounces

[David Cake]

At 6:21 PM +0200 8/9/04, Brad Knowles wrote:
>At 11:53 PM +0800 2004-09-08, David Cake wrote:
>
>>  	The only error I seem to be getting is
>>  dangerous permissions=42755 on queue directory /var/spool/mqueue-client/
>>  	which I can't seem how to turn off - I'm not even sure why its there
>>  are all, given that dir is not group or world writable - or how to turn
>>  off this warning with DontBlameSendmail
>
>	The problem could be with a parent directory above this one. 
>Or, it could be with a parent directory of a symbolic link pointing 
>to something in this path.  This can be a difficult one to debug.

	Oh, joy.
	Its not in /var/ or /var/spool, can't find anything in any 
place sendmail might be using that would link to here.
sendmail -v -d44.4 -bv listname claims its deliverable, and reports 
only a warning (which DontBlameSendmail should have dealt with)
check_perms is happy.
	I really am beginning to run out of ideas.


>>  (DontBlameSendmail? I want to go round to Eric Allmans house and slap him)
>
>	Everyone was riding Eric's case because there were so many 
>ways that people were finding to break into systems via weaknesses 
>that were not directly the fault of sendmail, but through which 
>sendmail gave them an attack vector.  He nearly killed himself 
>tightening down the security for version 8 sendmail so that this 
>sort of thing was no longer possible.

	Yes. I am well aware of what those settings means, and why 
they are called that. I have no urge to blame Eric for any security 
flaws that might result from using them. He must, at least, bear some 
responsibility for the extraordinary difficulty of diagnosing why 
sendmail won't do what it should, though.

>	Unfortunately, there are an infinite number of vendors who 
>ship an infinite number of systems that are themselves broken in one 
>way or another, and where the extremely strict security model 
>insisted upon by sendmail will cause other things to break.
>
>	That's why Eric came up with this option, so as to allow you 
>to shoot yourself in the foot (or blow it off with thermonuclear 
>weapons), if you so chose -- but he made sure that you would have to 
>explicitly configure sendmail to do that, and he made sure that when 
>the worst did happen as a result, you couldn't blame sendmail for 
>your security breach.

	Yes, indeed, I do not. But I can blame sendmail for being an 
overcomplex, incomprehensible, hard to debug, sanity damaging 
nightmare. Which I understand the historical basis for. But 
nevertheless, the frustration is very real, and aspects of sendmails 
design significantly contribute to that ongoing issue, aspects that 
Eric Allman has been known to deride himself.
	Don't get me wrong, I appreciate the pioneering efforts of 
those involved in the creation of sendmail (and .oz.au for that 
matter), but not having to live with certain dubious design decisions 
decades later.
	Just a couple more weeks and my last sendmail server moved to 
postfix. Can't wait. But will I ever get mailman working in the 
meantime.

>	If you want to tangle with me, I'll be glad to meet you in a 
>dark alley at an upcoming LISA or SANE conference.  Just let me know 
>when and where.

	Surely the appropriate when and where for that sort of thing 
is behind the bike sheds after school during junior high.
	Cheers
		David

... I felt very ambiguous about
creating something which I looked at and said "This is a monstrosity
by almost any reasonable definition." - Eric Allman
How did this happen? I should have gotten wise and said, "Wait! 
Something's wrong." - Eric Allman (referring to sendmail config file 
format)