Looping E-mails Beat The Net Down
Staili writes "Singapore-based women's magazine caused problems when it forwarded its mails to a large list of recipients, mainly mailing lists. In addition to security@suse.com, some help and subscribe lists were included; the type of addresses that tend to send out an automatic reply confirming receipt. And the loop was ready." I'm sure anyone who's messed with mail enough
has accidentally created a loop or two in their day, but this is really
slimey.
Its obvious that the women readers of the said magazine have the hots for German Linux developers and they tried to show their interest in them. True it wasn't in the best possible way but they did give a signal which the Suse guys completely misinterpreted. Sad.
At the intersection of computation and biology.
Somehow, few people seem to be able to get the autoresponder/autoforwarder thing right, despite the fact that it doesn't seem that hard and has been done correctly before. (Then again, there seems to be a dearth of good systems programmers around these days; I'm becoming increasingly cynical about such things.) Every day, I get auto-replies to MAILER-DAEMON's bounce messages, and every once in a while, some b0rken forwarder creates a mail loop. Unfortunately, when I try to tell the people responsible why what they are doing is a bad idea, they're usually not interested in hearing about the danger of mail loops.
Here are some things I've come up with over the years:
1) Never, ever auto-reply to MAILER-DAEMON or Postmaster (procmail has good regex macros for this -- use them or copy them).
2) Preserve the headers of messages you forward.
3) Set an X-Loop header and check for it (or *any* X-Loop header if you want to be paranoid).
4) Don't autoreply to the same address twice during [definable time period].
Those things just seem like common sense to me. Maybe someone else here knows more about the subject than I do. There has to be a HOWTO somewhere.
There seems to be some insidious 'oh, it's those clueless Asians' thread running through so many Slashdot posts recently that I think it's time the balance was addressed.
That thread is based on the emperical experience of thousands of mail admins throughout the world (not just the US, as your slashdot bash inaccurately implies). If those whose ISPs (and in some cases, countries) are being blocked wish to demonstrate otherwise, all they have to do is administer their mail servers competently and close down their open relays.
Until then, their inaction will speak louder than your words, be they from Singapore, Korea, or wherever. As one who has travelled to those places I am reluctant to block entire countries, but my boss doesn't want his mailbox filled with SPAM and if blocking half of Asia is how I appease him, then half of Asia will be blocked, period. My personal fondness of Asia (and, for that matter, Africa, and Europe, and other places I have had the privelege of visiting in the last several years) will play absolutely no role in this decision, and no role in my opinion of the (in)competence of ISP mail adminsitrators in those locations. The only metric of any concern is how many open relays there are, and how those responsible act (or, in the case of many notorious Asian providors, particularly in Korea, don't act) when the issue is brought to their attention.
As for the differences in phone systems, you are comparing apples and oranges, and assuming one causation (lack of technical knowhow) when a completely different causation (lack of well defined, enforcable government standards resulting from a lassaiz-faire market mentality in the last several administrations) is responsible, then trying to apply the erroneous conclusion derived from your erroneous assumption back to another issue that is, in any case, completely unrelated.
Internet booths are another example of the logical fallacy you have fallen into in making this argument. In a country in which more than half the homes have their own PCs, and just about every public library is already on the net (along with many schools), internet booths would be a profound waste of money. In other words, you have brought up another completely unrelated topic and misapplied it to your original argument, namely what approaches empower the most people to use the internet under what conditions, with those conditions in Singapore quite different from the United States, which in turn is very different from the UK or the rest of Europe. Clearly that has absolutely nothing whatsoever to do with the competency level of mail administrators in Asia, Africa, America, Antarctica, Mars, Pluto, the NGC-1 Nebula, or anywhere else for that matter.
The Future of Human Evolution: Autonomy
I once inherited a smallish network (70 nodes) that was using an NT box as a web gateway and mail server. It was running something called Xtramail, which is a truly bloody horrible piece of software. While I was trying to figure out how to gracefully get rid of this box (a 486 on ISDN), one of the users wanted to create a mailing list.
Ok, no problem. Read the docs, slurp this list, check these buttons, viola. One of the cute little checkboxes was "Only allow owner to send list mail." Duh - I checked it. The guy sent his email (only about 200 list members) and we went home.
I came in the next morning to 20,000 emails just in the queue. That fucker sent our tens of thousands of emails overnight, because the send restrict wasn't working. There were a couple dead addresses on the list, and they of course bounced - and Xtramail politely returned those bounces to the entire list. Wash, rinse, repeat. If that place had had a real server and a real 'net connection, it could have sent millions of emails in that time. As it was, many people on the list were (quite justifiably) pissed.
So I called up whoever owned Xtramail at that time (Artisoft at that time, but a different company now - can you say, "hot potato?") and had a slightly polite shit fit. The guy flat-out refused to acknowledge it was a problem, until I made him go through the same steps on his local copy.
Crickets.
"Uh, looks like that option isn't working. I'll have to file a bug report." Then I spent another 45 minutes trying to get accounting to refund the $200 I'd given them for the support call.
They never did fix the bug, but I gave up my plans to have a graceful transition. I pulled that POS out the same day and installed another little NT mailer, quite a nice one, until I replaced the whole thing with a qmail FreeBSD box.
No moral to the story, really ('cept I should have been more paranoid, and tested the list more). But I bet more than a few readers have had that quick "oh shit" feeling as they saw the queue filling up.
This isn't as much "normalization" as it is "don't take so many drugs when you're designing tables."