Re: ezmlm and delay notifies (was: Re: mini-bounce)
On Tue, Mar 16, 1999 at 10:08:27AM -0600, Bruno Wolff III wrote: On Mon, Mar 15, 1999 at 06:13:15PM -0500, Scott Schwartz [EMAIL PROTECTED] wrote: Peter van Dijk [EMAIL PROTECTED] writes: | But yes, it would consider these warnings as bounces. It also considers vacation messages to be bounces. :-( Vacation programs shouldn't be replying to lists. Lists should also identify themselves as lists so vacation programs don't reply to them. On this list in particular, when you subscribe, the ezmlm confirmation message doesn't include any of the magic cookies traditionally associated with daemon messages (such as "Precedence: junk" or "qmail-request"). My vacation program replied, and apparently that was enough to confirm my subscription. The confirmation process is a sham if it can be fooled so easily by vacation programs and autoresponders. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: Qmail mailing list and ReplyTo:
On Wed, Feb 17, 1999 at 11:52:22PM -, Russell Nelson wrote: Tim Pierce writes: Unfortunately, there's an awful lot of unreasonable mailers in the world, which makes that philosophy impractical. Pandering to the unreasonable mailers doesn't help. The chief cost is one of embarrassment to the poor slob who forgot that he was replying to a mailing list. We've all seen it happen. I can't imagine that anybody thinks that's a good thing. How's about we get the unreasonable mailers fixed? Sounds great! I'm all ears. Where do we submit bug reports for Microsoft Internet Mail, Microsoft Outlook, and WebTV? The sad reality is that mailers for the consumer world are getting worse and not better, and we have little power to fix that. Mailers that lack "group-reply" are only the tip of the iceberg; they also lack any useful filtering or filing capability, they fail to identify the message sender, they send replies to the wrong address, they send replies with broken return addresses. Managing a mailing list means making the decision about how to handle people like this. If you're a toy site, you can probably get away with telling all your users to lose the broken software. You can't get away with telling 50,000 users to lose their broken software. Pandering to these users doesn't necessarily help, but ignoring them is no better a solution. Ultimately you have to find some way to cope with their brain damage, until we figure out how to fix it. Like I said, I'm all ears. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: Qmail mailing list and ReplyTo:
On Wed, Feb 17, 1999 at 08:32:16AM -0500, Peter Green wrote: Why doesn't Qmail mailing list set the Reply To: field to "[EMAIL PROTECTED]". It is very anoying that I must type the mailing list address for every message I respond to. Check out http://www.unicom.com/pw/reply-to-harmful.html for a great reason *not* to set the Reply-To: header. Any reasonable mailer should have some sort of "reply to (l)ist, (s)ender, (b)oth" option. Unfortunately, there's an awful lot of unreasonable mailers in the world, which makes that philosophy impractical. While I sympathize with the opinions offered in "Reply-to Considered Harmful," it's mostly ivory tower theorizing. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: The ppiamdn annoyance
On Mon, Feb 15, 1999 at 11:19:13PM -, D. J. Bernstein wrote: These 43 useless messages from ppiamdn illustrate that unsolicited mail doesn't have to be commercial to be annoying. Note that majordomo's filters wouldn't have caught the messages. Gosh, it doesn't look like ezmlm did, either. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: 100,000 mailing lists
On Thu, Feb 11, 1999 at 11:43:39PM +1100, Mark Delany wrote: At 11:45 PM 2/10/99 -0800, Dongping Deng wrote: Let's consider a hypothetical situation: a machine needs to host 100,000 mailing lists, each list has subscribers, say, less than 15; and the traffic for each list is less than 3 a day. Lemme see. 100,000 * 15 * 3 = 4.5million deliveries a day. ... It's probably more appropriate to ask whether your underlying qmail system can deliver 4.5M messages a day. It's within the realms of possibility, but a standard single spindle system probably wont hack it. Is there a FAQ on configuring qmail to use multiple delivery queues? I can't see offhand how to do this from a stock qmail, and couldn't find any relevant info on www.qmail.org. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: concurrencyremote limit
On Thu, Feb 11, 1999 at 06:37:22PM +0800, Marlon Anthony Abao wrote: hello, with the release of the new linux kernel, the limit of concurrent processes is now raised. according to conf-spawn we cannot raise the qmail concurrency limit past 256. is there any reason for this? Qmail internally stores the concurrency limit in an object of type char. If you tried to specify a limit higher than 256, it would overflow and give you a concurrency limit *lower* than what you asked for. Is there a compelling technical reason why qmail shouldn't support more concurrent delivery processes, or is this just the result of short-sighted design considerations? -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: I don't trust 'em.
On Tue, Feb 02, 1999 at 05:22:47AM -0500, Cris Daniluk wrote: And what's to stop someone from buying a static IP from their ISP with its own lovely domain and spamming the world freely? The economics of static IP discourage it. ISPs in the U.S. often charge $200-300 in setup fees for static IP addresses, and typically an additional $100 per month. The spammer would have to be pretty sure that they would gross at least $400 per spam run in order to make it worthwhile, and I would guess that most spammers don't see anything close to that. Or relaying off of some server 2 thousand miles away that doesn't block relays? Some mail servers cant (for example sites like yahoo.com who have mail gateways... by the way, about 50-60% of spam I receive comes from "trusted" mail servers on mail gateways like this). More and more spammers are putting "ADV:" in their topics as is required by law and more and more are also sending "To be removed" messages. While the to be removed messages don't really work half the time, I think it is safe to say that a well constructed message filter could be made to block these out, if not on the MUA level, on the mail server level. In fact, our system-wide procmail filters include almost 200 recipes for blocking spam based on patterns in the message body. These include the Murkowski disclaimer, text like "hit reply to remove," "we are sorry if you have received this in error," "we are a responsible bulk emailer," "this is only an opt-in list," and other spammers' weasel words. We have a great deal of experience trying to block spam using full-text filters. The truth of the matter is that you can indeed stop a fairly high proportion of spam this way, but not enough to make it worthwhile to analyze the spam text and write new filters. Even 40% of a flood is still a deluge. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: Pattern-matching and filtering
On Wed, Jan 27, 1999 at 12:33:02PM -0500, Len Budney wrote: James Smallacombe [EMAIL PROTECTED] wrote: Selective filtering is ALL about pattern-matching. Correct, which is why it is flawed. If pattern matching were applied uniformly, then soon all spam will be 100% 822-compliant, and will originate only from hosts with valid MX records, and with exactly one envelope recipient and one envelope sender--which will be a valid email address. What will you match on then? The same thing I filter on now: body text. Want to know the absolute best ways of telling whether a message is spam? * Comes from a dialup: da.uu.net, dial-access.att.net, ipt.aol.com, as.wcom.net, etc. * Reference to "Section 301" or "Paragraph (a)(2)(C) of S.1618" * "This is a one-time mailing" * "If (I have reached you|you have received this message) in error" * "Authenticated sender" header in a message that doesn't come from Pegasus or Eudora. In principle, you are correct that this is an arms race, or that it could be. In practice, I find that these rules catch about 80% of the spam that moves into my systems, and the tweaks I have to place on my filters have become infrequent and insignificant. Even though I agree with you that it is not a workable long-term solution, in reality pattern-matching is a tool that I cannot live without. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: changing the VERP delimiter
On Thu, Jan 21, 1999 at 09:26:55PM -0800, Russ Allbery wrote: So if you just fix whatever it is that you're using to send mail so that instead of generating return addresses of the form: list-bounces-@host-@[] it generates them as: list+bounces+@host-@[] I believe you'll immediately get what you want. Mail is sent with a wrapper around qmail-inject, with an environment of: QMAILSUSER = list-request QMAILSHOST = rootsweb.com QMAILINJECT = r Am I doing it the wrong way? This is the only reference to VERPs I could find in the qmail-inject, qmail-send, qmail, qmail-smtpd etc. man pages. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
Re: changing the VERP delimiter
On Fri, Jan 22, 1999 at 03:01:36AM -0800, Russ Allbery wrote: D J Bernstein [EMAIL PROTECTED] writes: Yes. Dash-separated extensions are used in the .qmail-*-default mechanism, qmail-inject VERPs, ezmlm VERPs, etc. conf-break is the default user-ext delimiter. It doesn't affect the use of dashes inside extensions. Am I correct in thinking, then, that the "right" answer to Tim's problem is not the patch I provided but rather to use a QMAILSUSER of list-request+bounces instead of list-request, so that sendmail will still deliver to list-request and the dashes instead of + won't matter? That does look like the right solution for me. Thanks for the idea. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
changing the VERP delimiter
[ I sent this to qmail-help a month or so ago, but had no response. ] I'm using qmail as the outbound mail agent on a machine that runs sendmail for incoming mail. I would like to modify qmail to use "+" in constructing per-recipient VERPs on outgoing mail. That's necessary to make sendmail accept the bounces, and would permit me to hack SmartList to take advantage of VERPs for accurate bounce processing. I thought that changing conf-break would change the character used to construct VERPs. However, it looks like conf-break only affects the delimiter that qmail-smtpd looks for on incoming messages. No matter what's in conf-break, VERPs are still constructed with "-" as the delimiter. Is this intentional? It doesn't seem to make sense to me, and I'd like to know what I'm missing. -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades -- Regards, Tim Pierce RootsWeb Genealogical Data Cooperative system obfuscator and hack-of-all-trades
