ietf-smtp
[Top] [All Lists]

Re: Question on retry strategy in 2821

2007-06-13 15:47:36

Rolf E. Sonneveld wrote:

Hi,

today I got the following mail from Postmaster at getronics.com (Getronics being an ICT company with many thousands of employees, spread over several countries):

<quote>
The sending system (IP address <my MTA's IP address here>) apparently retries delivery only every two hours. The intial retry after a 421 status should be less than 30 minutes. Please refer to RFC 2821, section 4.5.4.1 "Sending Strategy", where the recommended first retry should be 30 minutes, and there should be two retries within the first hour. The initial retry after a 421 status should probably be somewhat shorter (say 15 - 20 minutes).

The problem here is that the greylisting mechanism used by Getronics remembers the connection attempt for only one hour. After that the next connection is seen as if was a new, and not a retried, connection and it's greylisted again.

The owner of the sending server should adjust the first two retry intervals (and certainly the first one) to fall within the recommendation made in RFC2821 to avoid problems when interacting with other email systems.
</quote>

Am I misinterpreting RFC2821 in that (IMO) there is no such requirement as stated here?

It is not my intention to start a (flame) war on greylisting, but keeping greylist information for only one hour needs a reality check, isn't it? I know at least two mature and well-known MTA implementations that have a default retry interval which doens't match the requirements of this domain.

/rolf

Right. These are no requirements but suggestions. But RFC2821 does suggest at least 30 minutes.

4.5.4.1 Sending Strategy

   .... In general, the retry interval SHOULD be at
   least 30 minutes; however, more sophisticated and variable strategies
   will be beneficial when the SMTP client can determine the reason for
   non-delivery.

IMO, if you have incorporated a greylist system or wish to consider greylist encounters with outbound mail, then you really do need an variable frequency, with a much shorter retry for at least the 2nd attempt. Otherwise, greylist will give you a blackeye. Greylist is real. People use it, and statistically it was found that at least 1 hour (or < 60 minutes like 55 mins) is used as the hold/block time.

We use to have a DEFAULT 1 hour retry, 72 attempts or 3 days. Sysops changed it as they wanted it. But the frequency was fixed per attempt.

When we added greylisted, it is immediate and obvious that a more variable frequency table was prudent.

Today we give sysops a frequency table and the default table is:

[Attempts]
Default=60
Attempt1=5
Attempt2=5
Attempt3=15
Attempt5=30
Attempt10=120
Attempt21=5
Attempt22=5
Attempt23=15
Attempt25=30
Attempt30=120
Attempt40=60
Attempt72=60


--
Hector Santos, CTO
http://www.santronics.com
http://santronics.blogspot.com

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