Great Circle Associates List-Managers
(October 1995)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: Allowing unbracketed IP addresses?
From: Keith Moore <moore @ cs . utk . edu>
Date: Tue, 03 Oct 1995 12:32:06 -0400
To: Keith Pilotti <kfp @ qualcomm . com>
Cc: List-Managers @ greatcircle . com, moore @ cs . utk . edu
In-reply-to: Your message of "Mon, 02 Oct 1995 18:22:51 PDT." <199510030122 . SAA19884 @ philly . qualcomm . com>

> I'm having a debate with one of the admins at our site who wants to allow 
> unbracketed IP addresses on our lists, and to modify sendmail's config to 
> process them.
> 
> This type of address is frequently produced by newbies and other clueless 
> types with misconfigured mailers, and the other admin does not want to 
> "exclude" these people from participating in our lists.
> 
> I see this as prologing the agony of newbieness on the Internet, he sees 
> it as "being friendly".  (I'll let you guess which one of us has 15+ years 
> on the net and which one has 5. :)

I'm not sure what you mean by "unbracketed IP addresses on our lists".

If you mean that you will accept subscriptions for people with malformed
email addreesses, I recommend against doing so.  

I used to do this, but I found that such addresses almost invariably started
bouncing messages fairly soon after subscribing (sometimes even beforehand).  
After I put a syntax checker in my mlm that refused a subscription from any 
invalid address, a major source of my bounce messages went away.

However, the routine that bounces invalid does attempt to guess the
person's correct address, and returns that guess as a suggestion.
(for example, if the address was user @
 uk .
 xxx .
 yyy .
 zzz, and domain
uk.xxx.yyy.zzz was found to be invalid, it guesses user @
 zzz .
 yyy .
 xxx .
 uk;
if the address was user @
 xxx and the message came from domain
xxx.yyy.zzz, the guessing routine will suggest user @
 xxx .
 yyy .
 zzz)

This seems to work - there are many instances in my logs where an 
initial attempt at subscribing fails because of a bogus address but
a subsequent attempt succeeds.

IMHO, we're not doing users a favor by accepting invalid input and 
trying to interpret it, because (a) this just encourages users to
generate invalid input (after all, it "works" ... so they'll tell
their friends) and (b) because well-formed input has a well-defined
interpretation but invalid input does not -- and sometimes we'll guess
wrong.

Finally, if you do want to accept user @
 ip-address anyway, you don't
want to do it in sendmail.  Instead, teach your list manager to
rewrite user @
 ip-address to user @
 [ip-address] .
   I could go into
great detail about why modifying sendmail in this way is a bad idea, 
but I'll just say that it's fairly difficult to get sendmail rewrite
rules to tell the difference between an ip address and a domain, 
and any changes you make to a sendmail config file have the tendency
to have unwanted side effects.

Keith Moore


References:
Indexed By Date Previous: Re: Mail delivery software [was Best List Software for SCO]
From: Matti Aarnio <mea @ utu . fi>
Next: "price" of receiving mail
From: cathy @ ncbapsun2 . pet . wfu . edu (Cathy Eades)
Indexed By Thread Previous: Re: Allowing unbracketed IP addresses?
From: Paul Haas <paulh @ hamjudo . com>
Next: Re: Mail delivery software [was Best List Software for SCO]
From: Matti Aarnio <mea @ utu . fi>

Google
 
Search Internet Search www.greatcircle.com