spf-discuss
[Top] [All Lists]

Re: Latest proposal re HELO checking: make HELO tests optional

2004-03-12 03:17:19
On Fri, 2004-03-12 at 01:45 -0800, Greg Connor wrote:
You are quite right, I didn't review the message in the context of previous 
ones, and therefore jumped to conclusions.  Please excuse... those remarks 
were intended for someone whose secondary just accepts all mail all the 
time :)

NP. You were just following the lead set by others; I appreciate that
there _are_ people who think what you thought I thought, and they do
genuinely deserve a LARTing. Highlighting the assumption you thought I
made, and pointing out that it's broken, is potentially useful for the
peanut gallery since they may also have missed the same context you
missed.

Agreed.  Do you think it would be that bad to not have a secondary at all? 
Or to have the secondary give a 4xx result if the address can't be 
verified? 

Those two are basically the same thing, at least in the case that the
address is verified by callouts. It's not the end of the world to have
no secondary, certainly. However, I find it extremely useful to be able
to collect pending mail into one place and flush it with ETRN, rather
than telling the users "Oh, that urgent mail you're waiting for will
probably get resent by the end of the day". It's a quality-of-service
issue.

I also consider it quite impolite when subscribers to my mailing lists
have frequent outages and no MX backup, because their mail ends up
sitting on my queue for long periods of time. I have been known to kick
people off the lists for no better reason than they have no MX backup
and their primary has been down for a while.

 How long does it take to receive the queued mail after the 
primary comes back up? 

I have no way of knowing the answer to this, because it could be queued
anywhere. I know the answer is less optimal then "when I flush the
queue", certainly :)

 All are things to consider when weighing whether to 
have a secondary.

Indeed. I prefer to have a secondary. But if I couldn't set it up in
such a way that accept-then-bounce is _extremely_ rare, I'd do without. 

I'm satisfied with 'extremely rare' rather than 'impossible' in the
above sentence. Wayne's opinion seems to differ; that's his right.

But yeah, accepting then dropping is bad.

I'd certainly never opt for accept-then-drop. That would just be broken,
IMHBCO. Wayne and I appear to differ on that preference too.

Some of the worst offenders were sending us 1000 bounces a day... so by 
going to 454 we will be increasing that to 1000 every 15 min for the first 
day, and up to 5000 per 15 min the fifth day.  That ought to get someone's 
attention.  

That's just scary. Personally, I'd be reporting these to the abuse
contact at their network provider. They are contributing to a DDoS and
should be treated as such.

Hmmm, thinking of sender verification -- some people do callouts but
reject the incoming mail only if they receive a _permanent_ error when
pretending to deliver a bounce. They'd actually be accepting mail
claiming to be from *(_at_)altavista(_dot_)com now you switched to 4xx from 5xx.

-- 
dwmw2



<Prev in Thread] Current Thread [Next in Thread>