1997-01-20 - Re: Sandy and the Doc

Header Data

From: Bill Frantz <frantz@netcom.com>
To: cypherpunks@toad.com
Message Hash: 5e166da6b59d9ee5591f7c63f12debf512c28ec22a2f32557110137b851134b4
Message ID: <199701202142.NAA19130@toad.com>
Reply To: N/A
UTC Datetime: 1997-01-20 21:42:52 UTC
Raw Date: Mon, 20 Jan 1997 13:42:52 -0800 (PST)

Raw message

From: Bill Frantz <frantz@netcom.com>
Date: Mon, 20 Jan 1997 13:42:52 -0800 (PST)
To: cypherpunks@toad.com
Subject: Re: Sandy and the Doc
Message-ID: <199701202142.NAA19130@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


At 9:19 PM -0800 1/19/97, WinSock Remailer wrote:
>Bill Frantz <frantz@netcom.com> writes:
>
>> I hope if such a system is set up, people who use it will realize that
>> email is not 100% reliable.  Just because algebra.com sent mail to
>> toad.com, doesn't mean that toad.com actually received it.  A small
>> fraction of a percent of these messages will be lost.
>
>Failing something catastrophic like a disk failure or a host down for
>more than a week, this should not be the case.  Almost every site
>running sendmail has the Os ("SuperSafe") option set.  That means
>sendmail will not respond to a "." at the end of a DATA command with
>SMTP code 250 until it has written the incoming message (and queue
>info) to disk and called fsync.  Thus, you may get 2 copies of a
>message, but mail messages should not just disappear regularly at all,
>even if the network goes down or a machine crashes.
>
>Non-sendmail MTA's tend to be even more strict about this, not even
>allowing this behavior to be disabled.

About 9 months ago, I lost somewhere between 50 and 100 email messages
because they were written to a disk with a corrupted file system.  Now
failures of this kind may be in the category of "catastrophic disk
failure", but in my unfortunate experience, they are not all that uncommon.

Note that since I lost multiple messages due to one failure, tests of the
form, "Once is happenstance, twice coincidence, three or more is
conspiracy" might say, "conspiracy" even for just one failure.  The chances
of this test criteria error increase for frequent posters.

At 4:02 AM -0800 1/20/97, Toto asked:
>  Are there any critereon established, yet, as to what type of content
>will be necessary for a letter to get 'lost'?

See the above discussion of disk failure.


-------------------------------------------------------------------------
Bill Frantz       | Client in California, POP3 | Periwinkle -- Consulting
(408)356-8506     | in Pittsburgh, Packets in  | 16345 Englewood Ave.
frantz@netcom.com | Pakistan. - me             | Los Gatos, CA 95032, USA








Thread