Fog Creek Software
Discussion Board




Anti-spam vigilantes

The "filtering good mail" spam problem is significantly worse than Joel metions in his article.

We run a business that generates a lot of email.  We regularly get spamcop complaints (and automated complaints from other spam filter software) about confirmation email (i.e. confirmed identity for a double opt-in list), which is obviously not spam by anyone's definition.

We generate so many confirmations, and so many bogus complaints, that we've had problems with various Internet services (DNS providers, ISPs, etc) that have a "zero tolerance" spam policy.  Over the course of nine months, we've had four significant service interruptions caused by a "guilty until proven innocent" attitude.  The state of the Internet today is such that less than a dozen complaints can get you pulled off the Internet by these vigilantes, with no warning.  Careful reading of their terms of service says they reserve such rights.  There are remarkably few service providers that don't have such wording in their terms of service.

So, in addition to the usual crackers, DDoS script kiddies, etc., something else many large web sites have to worry about are overzealous system administrators trying to save the world from spam.  I can't wait until the script kiddies figure out they can target a DoS attack by just sending bogus spam complaints...

Anon
Monday, November 18, 2002

I disagree.

You have to decide up front just who the internet is meant to serve -- you, the advertiser/commertial communicator/whatever, or me, the reader.  Obviously, I choose me.

I have to make a decision when I set up a mail address -- just how much noise am I willing to tollerate, and how much signal am I willing to lose in noise-supression attempts.  If I decide to use a service which may or may not identify your inbound mail as possible spam, that's entirely up to me to decide.

Joel's points on the tag are well taken, but the problem there is that one still has to sit around while the tagged spam is downloaded and sorted -- not a huge undertaking on a corporate network or a high-speed connection, but it can be murder on dial up connections.

David Mackintosh
Monday, November 18, 2002

Totally disagree. I can tolerate noise. I absolutely cannot tolerate missing signal. This is the major deficiency with virtually every spam blocker. For some reason they don't understand that blocking legit emails is a non-starter.

pb
Monday, November 18, 2002

So, you, the reader, decides to set up a mailing list for your favorite topic.  You follow the rules, and use software that operates a double-opt-in list (they request to sign up, you send them an email to the address they give, they reply and are signed up). 

However, your favorite topic is the plight of Tibet, so you set up a mailing list on 'freetibet.org'.

90% of the people trying to sign up for your service can't, because their fully automated anti-spam software that came with their mailer software and about which they are largely unaware, blocks any email from you because the word 'free' appears in most spam, and therefore your confirmation email must be spam.

Does this serve you, the reader?

Or, you set up a website on 'freetibet.org', and someone who takes a dislike to your site generates a set of bogus spam complaints to your registrar, your DNS provider, your ISP, your ISP's ISP, etc., and one of those people decides to implement their companies zero-tolerance policy and removes your service (really appealing when the registrar deletes your DNS record with no warning, and you lose freetibet.org to someone else who's been camping on that address).

How does this serve the reader, exactly?

Anon
Monday, November 18, 2002

That would be free tibet FROM CHINA!

obvious
Monday, November 18, 2002

For every one "freetibet''s their will be a thousand (or more) "freecreditcards" or such like. So it's still worth it. Its very unfair on the "freetibet" people admittedly, but using your example, they can change their name, so it's no more than an inconveniance for the greater good, like wearing seat belts.

FreeMeFromSpam
Tuesday, November 19, 2002

I'm very sympathetic to false positives.  In my opinion, if even one legitimate mail is lost due to anti-spamming techniques, that's a real problem especially to a business.

The way we handle it is to have all mail marked as spam go to a mailbox and every morning go through that mailbox and manually inspect the messages.  For the most part it's really easy and quick to figure out what's legitimate (or in some cases, possibly legitimate) and not legitimate by just looking at the subjects.  I've been doing it for several months now but I'll soon be passing it back to IT.  It takes me 1/2 an hour each morning to go through about 700 messages.

The way our company looks at is that it's a lot cheaper for one person, skilled in groking good/bad messages than to hand that off to the rest of the employees and cause them to get distracted, lose productivity, etc.

Chris

Chris Blaise
Tuesday, November 19, 2002

How about a confirmed opt-in for anti-spam software.

That is to say that any mail server that wishes to apply any spam mechanisms sends me an email explaining that it is intending to block any email to me that comes from a list of unilaterally blacklisted servers or contains any reference to freedom (political, software, judicial or otherwise) or anything else that it feels like, and could I please sent them a reply stating that I am in agreeement with them deleting my messages or otherwise they will either send me all my mail, or refund my subscription and promise to forward on all my email to my new provider for a period of a year so that I may ensure that everybody has the new address.

Also Microsoft could have the confirmed opt-in for SP1 for IE6 and Windows XP which blocks completely a long list of attachment types. It should state that the service can only be turned on after I have sent them an email confiriming my wish to have them block possible viruses as well as software updates and messages and photos from my friends and family.

Stephen Jones
Tuesday, November 19, 2002

"That is to say that any mail server that wishes to apply any spam mechanisms sends me an email explaining that it is intending to block any email to me that comes from a list of unilaterally blacklisted servers or contains any reference to freedom (political, software, judicial or otherwise) or anything else that it feels like, and could I please sent them a reply stating that I am in agreeement with them deleting my messages or otherwise they will either send me all my mail, or refund my subscription and promise to forward on all my email to my new provider for a period of a year so that I may ensure that everybody has the new address."

So you are saying you can tell me how to run my server?

I don't like the "Anti Spam vigilantes" nor do I like the "my free speech first, on your property" vigilantes. I like running my little network how i choose. If I choose to stop mail crossing my network for *any* reason thats up to me, as long as I am up front with my users about it.

I'm not about to pander to the needs of either side, I just want my network to work.

Robert Moir
Tuesday, November 19, 2002

Dear Robert,
                    So if the mailman decided to throw away your letters you would say that's OK because the mail bag is his property. No problem if an airline taking you from New York to Australia drops you off at Mogadishu because it's their plane and they don't like getting told what to do with it?

                      If you are an ISP you are promising to deliver me mail. If you don't want to then advertise that loud and clear (We are an ISP that only delivers the mail messages to you we feel like and dumps the rest) and then give sufficient notice for those clients who wish to change to another without such illusions of grandeur.

Stephen Jones
Tuesday, November 19, 2002

"                  So if the mailman decided to throw away your letters you would say that's OK because the mail bag is his property. No problem if an airline taking you from New York to Australia drops you off at Mogadishu because it's their plane and they don't like getting told what to do with it?"

Sure if they told me up front they might do that. Of course, if they told me up front they might do that, I could well take my business elsewhere, but if i continued to use them knowing they might do that to me, I think my options would be limited if they decided to exercise that right don't you?

And your reply is somewhat disingenious anyway. No one is talking about randomly throwing mail away, they are talking about deciding that they don't want to exchange email with certain other networks for various reasons.

Again, if i am honest and up-front with my users about what i am doing, why, and what criteria i use to decide what mail to keep and which to reject then what is your problem really?

If I decide to reject email from everyone called "Simon" that might make me the worlds dumbest email administrator but if I am totally up front with you and make sure you know thats what i do, then if you still choose to allow me to manage your internet access then thats up to you.


"                      If you are an ISP you are promising to deliver me mail."

You really need to read the terms and conditions of your agreement with your ISP. Because unless you use the stupidist ISP in the world, they "promise" nothing of the sort, I'm sure.

Every ISP I've worked with has only ever promised to ensure its own service is in working order to a certain SLA, and to make "best efforts" - that's all you can do because you cannot guarantee delivery of anything on the internet with 100% certainty for lots of reasons that have nothing to do with spam filters or dumb network admins.

They might use a lot of flowery language to make you think that in the event of a mail server crash the ISP's CEO will personally print off your email and fly around the world hand delivering it, but if you read closely, I am sure it will add up to "best efforts". Go on, check now. I'll wait. Back yet? What did it say? See, what did I tell you. Moving on...

" If you don't want to then advertise that loud and clear (We are an ISP that only delivers the mail messages to you we feel like and dumps the rest) and then give sufficient notice for those clients who wish to change to another without such illusions of grandeur. "

As I said in my original post, "As long as I am up-front with my users". So did you have a point or are you just flapping your lips for the sake of it?

Go on, tell me I'm violating your constitutional right to something or other you once read about on slashdot, I know you want to.

Robert Moir
Tuesday, November 19, 2002

I wouldn't ever work with snail mail if I were you. In Europe if you threw the mail away and were stupid enough to be upfront about it you'd end up in jail because it's a criminal offense. I suspect the same is true of the States.

I never said an ISP guaranteed to give me mail. I'd accept best effort, but deliberately refusing to deliver mail is not "best effort". And don't come along and say that all your doing is choosing what networks you do business from. If you are a mailman you accept all mail that is going to a particular address, bombs and perishable goods aside. You can't say you don't want mail from Alabama becaise it's often wrapped in pretty paper that attracts the flies in summer.

And your rubbish about being upfront is just a smokescreen, because the whole point about my post was that I would accept an ISP that was up front. All it has to do is tell me it won't deliver some of my messages, give me a month's notice to find another provider, reimburse me any money I've lost by the change, and then ensure that mail sent to the adress they've said they will deliver to continues to come to me for the time it takes for all my contacts to get the new mail address. Of course the ISP won't do that because it wouldn't have any customers.

Offer the Spam blocking as a service if you want, but make it an option. We're not talking about nebulous rights here, we're talking about a heck of a lot of money being lost by messages not getting through. If somebody sends an email offering a contract or a job, and that email doesn't get through because your server has decided to bounce it, then that one bounced email has cost tens of thousands of dollars. It doesn't take many cases like that to cost more than the total worth of your mail server business.

Stephen Jones
Tuesday, November 19, 2002

"If you are an ISP you are promising to deliver me mail."

"I never said an ISP guaranteed to give me mail. "

Both quotes from you. Make your mind up. Its pointless trying to have a discussion from you until you can remember from one post to the next what your own position is.

"In Europe if you threw the mail away and were stupid enough to be upfront about it you'd end up in jail because it's a criminal offense. I suspect the same is true of the States."

- I wouldn't know about the states, as I live in Europe.

"If you are a mailman you accept all mail that is going to a particular address, bombs and perishable goods aside. "

Some would argue that spam is an internet bomb. I'm not sure I'd go /that/ far myself, but the people who do go that far have plenty of things they use to justify it.

Do you know AOL won a court case over spam and submitted evidence that said spam had a direct impact on their business to the extent that a couple of dollars per month of each user's fees was directly traceable to resources needed to keep their systems going under the workload of all that spam.

Add in mass mailing worms too and you can easily justify spam on a computer as being in the same frame of mind as letter bombs in the post.

Robert Moir
Wednesday, November 20, 2002

There's a difference between promising and guarranteeing. I'm not being in the least inconsistent. We both agree on "best effort"; it's just that you have a very strange idea of best effort that includes making considerable effort to ensure that mail doesn't reach the sender.

To make stupid comparisons between an internet spam and a letter bomb is puerile. Snail mail has the exact equivalent of spam, which is called junk mail, and mail companies actually fight over the right to deliver it.

We are not talking about the right to receive spam; we are talking about the right to receive genuine messages that are blocked because whole bunches of URL's are blocked.

To the best of my knowledge AOL was suing a spam mailer and I only wish a few more ISP's would do so, I would willingly contribute a couple of dollars a month or more together with any time needed to the cause. However I don't think many of AOlL's or any other IDP's customers, would be happy to have a dollar or two a month returned to them in exchange for accepting the occasional deleltion of genuine messages that could cost anything from a few tens of dollars (missed trip to the airport, telephoning a relative abroad for news because his emails didn't get through) to several thousands of dollars (missing a job offer or business contract) for each one that didn't get through.

I still maintain that there are ways of cutting down at spam without putting your clients' communicablity at risk. Why not keep a single copy of every mass mail you bounce, and every day, or every week, send an automated log tile to each customer asking him if he wants any of the messages. This would also be an excellent way of enabling you to build up a white list, so that Joel's newsletter would get through but the commercial junk that WhatCounts.com tacitly encourages doesn't.

Stephen Jones
Wednesday, November 20, 2002

"There's a difference between promising and guarranteeing. I'm not being in the least inconsistent. We both agree on "best effort"; it's just that you have a very strange idea of best effort that includes making considerable effort to ensure that mail doesn't reach the sender."

I don't make "considerable effort" to block email at all.

"To make stupid comparisons between an internet spam and a letter bomb is puerile. Snail mail has the exact equivalent of spam, which is called junk mail, and mail companies actually fight over the right to deliver it."

Spam and postal bulk mail are not exact equivilants, and the reason i say that is the reason mail compaines fight to deliver it. When someone sends you bulk snail mail, THEY pay. When someone sends you spam email, YOU (and your ISP and those places along the route) pay.

Quite a difference. I'm betting most post services do indeed reserve the right to  refuse to deliver mail shots that haven't been paid for.

"I still maintain that there are ways of cutting down at spam without putting your clients' communicablity at risk."

Yes there are, but are they viable? I'm not trying to be insulting but your suggestion below suggests you either have never run a mail server, or at least that you've never run one with a very high volume of messages.

" Why not keep a single copy of every mass mail you bounce, and every day, or every week, send an automated log tile to each customer asking him if he wants any of the messages."

Ohhh because of the prohibitive disk storage and other resources this would use up, for just one reason. The problems with users having trouble working out which ones are signal and which are noise, resulting in them either ignoring the list (and losing mail) or downloading everything and getting spammed to death.

" This would also be an excellent way of enabling you to build up a white list, so that Joel's newsletter would get through but the commercial junk that WhatCounts.com tacitly encourages doesn't. "

Thats a fair point.

Incidentally, I want to make it clear - I don't like block lists. I do, however, think they are the lesser of two evils. 

It's also wrong to assume all block lists are born equal; there are some which (attempt to) only list servers that are actually currently sending spam. Assuming an ISP told it's customers that it was using a list like this, then such a list MIGHT be suitable for an ISP.

There are some very aggressive block lists which block IP ranges if spam repeatedly comes from anywhere in that area. My current employer is in a position where we use this block list (we tell our users that we do) because it makes sense for us. I'd say this list was *not* in the slightest little bit suitable for general use.

We also use our own blocklist. That obviously does what we want. It also proves the blocklists do not, entirely, or we wouldn't need this as well.

The problem, for me, is not that people use blocklists, but that some people use blocklists without understanding fully what criteria is being used by that list and hence get results they did not expect.

Robert Moir
Wednesday, November 20, 2002

I love the "prohibitive storage costs" that mail servers throw at you.

Storage at the moment costs something like two cents a megabyte, using EIDE disks. You can factor in a lot of other things but the cost is still low.

I'm not suggesting you keep every message that comes through, but surely there must be a way to run a checksum and just keep one copy of a mass mailing.

I can't help feeling that the reason things like this are not done is that the system administrator, whether of an internal or external mail server, doesn't really think it matters a great deal if the users don't get their messages.

To go off slightly on a tangent, the system admins who posted on another forum defending their rights to block Word docs or zip files from coming into the LAN "because nobody really needs them"  have the same desire to rule the world; they should be told to run for President or given a copy of the Sims.

Sure, blocking .docs and zip files makes the sysadmins job easier. It also halves the efficiency of your HR department wihich end up having to advertise vacancies with a hotmail address (and of course lose half the resumes because of the 2MB limit).

Blocking mail has an effect on the user; before you do it you need to consult with him and remember that if he says he needs something he's going to be right.

Stephen Jones
Wednesday, November 20, 2002

"Storage at the moment costs something like two cents a megabyte, using EIDE disks. You can factor in a lot of other things but the cost is still low."

What if I don't use EIDE drives, which as you say are cheap, but I use SCSI raid arrays to store mail on? Thats not so cheap. Also, if I store it I have to back it up. Modern high capacity tapes and tape drives are more expensive than you probably think.

"I can't help feeling that the reason things like this are not done is that the system administrator, whether of an internal or external mail server, doesn't really think it matters a great deal if the users don't get their messages."

Then you'd be mistaken, imho. I've never met someone who considered their job to be email server administration who felt that way.

"To go off slightly on a tangent, the system admins who posted on another forum defending their rights to block Word docs or zip files from coming into the LAN "because nobody really needs them"  have the same desire to rule the world; they should be told to run for President or given a copy of the Sims."

If they were blindly doing so regardless of user requirements then I'd agree with you. However take the example of a network that is used by a private business, not an ISP, which uses unix based systems for all it's users. Word doc and windows zip files are of little use to them and they may have some justification in taking that action.

"Sure, blocking .docs and zip files makes the sysadmins job easier."

How so? If I set up a mail server, and configure it to exchange email with hosts on the internet, then to further configure it to accept all kinds of attachments is in fact the "easier" choice; it's most likely that this is the default option and i'd have to invest time and effort in blocking those attachments. So how is doing something that causes me more work the "easy" choice?

"It also halves the efficiency of your HR department wihich end up having to advertise vacancies with a hotmail address (and of course lose half the resumes because of the 2MB limit). "

Not if said HR department say to applicants: "we only accept resumes in text format, rich text format, or html" or some other nonsense.

A surprising amount do just that you know, including some (Microsoft spring to mind) who want your resume as a word document and will get upset if you use anything else.

Draconian? Maybe. But then they already expect your resume to be laid out in a certain way and probably won't look at it if you use an unconventional layout, and they probably expect you to submit job applications on a schedule thats conveniant for them rather than you, etc.

If the HR department don't use Word and cannot open word documents (I'm still at my example unix based company), then what benefit would allowing them attachments they can't open serve?

"Blocking mail has an effect on the user; before you do it you need to consult with him and remember that if he says he needs something he's going to be right. "

I've always said this. I don't know why you persist in your fantasy that I didn't. You might remember that some people may find blocks have a positive affect on their email. The fact is, there are no easy answers to this dilemna.

Robert Moir
Wednesday, November 20, 2002

"The fact is, there are no easy answers to this dilemna. "

Thait is what we are agreed on. What this whole thread is about though is differing priorites. For you and the poster on the "spiel" thread the priority is avoiding spam, whilst for me and the original poster on this thread it is ensuring that I receive all my legitimate mail.

I've found myself sending out details to job candidates that never reached them because our college mail server was on a black list because of an open relay. This went on for at least a month, with some messages simply not being delivered, some being returned as undeliverable, and finally one coming back with an explanation. I checked out all the listing sites, sent off the info to our sysadmin and when he had finally understood what an open relay was we managed to get ourselves off most, and maybe all, of the blacklists.  Now it is more than possible that one of the people we mailed went and accepted a job at half the salary because he thought we had discounted him. Over a couple of years we are talking of that guy having lost twenty to thirty thousand dollars.

The discussion about blocking word docs and zip files was one held on the ZDNet forums a year ago. There were only two of us who defended the necessity of not having them blocked, or stated that the policy of not ever opening attachments from unknown senders was stupid, impractical and considering the provenance of most viruses ineffective. Still their policy maybe explains why systems engineers never manage to get a date. They refuse to open the conversation with anyone who hasn't opened it with them first!

On another tack, what advantage is there for using SCSI disks over EIDE disks, now that RAID for EIDE disks is standard. The reason is I'm speccing out a network for a new training centre we're opening next September, and  I'm very tempted to have four identical single processor EIDE servers so I can make backups with cloning software and if the server goes down and I can't use the mirrored copy, simply slamming the clone on to the fifth spare server. if one or two of the machines were SCSI I would need a second back up machne, or a different strategy.

Stephen Jones
Wednesday, November 20, 2002

Yeah lets move on :-)

Well the big advantage of EIDE is cost, as you know. One problem is that right now, a lot of EIDE RAID cards that do anything other than the absolute basic minimum tend to be rather expensive, which nukes some of the price benefit. Unless of course the absolute bare minimum the entry level EIDE raid cards do is just fine for your needs. From what you describe of your needs I'd say this may well apply.

As for performance, the gap here has certainly closed, which does mean EIDE drives for servers are viable whereas that wasn't the case a while ago. Where SCSI (and therefore SCSI raid) still scores is in sustained data throughput speed, rather than burst speed, and keeping performance good with multiple drives on the same channel. I'd say this wasn't a big deal for a file and print type server, but if I was designing a big SQL database server, or an Exchange server or similar things to either of them, I'd specify SCSI RAID because it would cope better with sustained load.

Lastly, a lot of IDE drive vendors only guarantee IDE drives for a year, whereas SCSI is guaranteed for 3 years still. You've got to wonder why. And if reliability of the individual drives on your system is a big deal, the answer I come up with when I ask "why" doesn't do much for EIDE's cause.

Robert Moir
Thursday, November 21, 2002

The file and print server is going to be EIDE, obviously.

There will be a SQL database server though the database will not be that big (basically marks, attnedance and scheduling for a maximum of 1600 students and 100 staff, though probably much less). Databse size will be smakl, and I would want RAID for mirroring not striping. The users (staff only) would probably only be connecting for ten minutes a day, though a lot would be connecting at the same time when lessons finish.

What about Exchange Server? We would have one for the staff (maybe a total of 120 users with staff and admin) although again it would not be used a great deal. With all the mail kept on Exchange mailboxes what kind of setup should I be thinking of?

Stephen Jones
Thursday, November 21, 2002

Exchange again is just another database, you can size servers for it by thinking of it as just another SQL database (It runs JET at the moment rather than SQL but the concepts work regardless).

For low use for a small amount of users then mirrored IDE or SCSI, your choice, would be just fine. If money is tight, getting 4 fast IDE drives and a good IDE raid controller and having two mirrors, one for operating system and transaction logs, and the other mirror for the database files themselves, would probably give equal or better performance than trying to squeeze SCSI raid 5 out of a tight budget and ending up with low end SCSI stuff.

Hope that helps.

Robert Moir
Friday, November 22, 2002

Thanks Robert,
                        This is Saudi, so the problem isn't so much not getting enough money as stiopping them wasting a small fortune (and then cutting of the oxygen because they've spent so much).

                          The parent college, with coming on to 3,000 students and over 200 staff has as part of its multi-million dollar ATM network a $35.000 four processor 550Mhz Pentium Xeon Compaq server, among others. This server runs at between 5 - 8% CPU time. The trouble is trying to persuade people that a server is just another computer, with slightly different requirements than a workstation. They keep coming up with ideas like "future-proofing"; I keep tellling them it's the equivalent of buying a Greyhound Bus when you get your first job out of college so that eventually you'll have room for all the great-grandchildren, even though the bus will have rusted away before your kids even grow up and when you get to the age to have grandchildren they'll be giving Greyhound Buses away with a litre of gasoline,

"Suppose the server isn't powerful enough?" "Then  give it to the receptionist and buy another. Total loss under $500"

On an unrelated front, how is it that people keep claimng that Access is not robust enough for more than about five users, when it has the same database engine as Exchange Server which deals with hundreds or thousands of users (albeit it very badly).

Stephen Jones
Friday, November 22, 2002

"On an unrelated front, how is it that people keep claimng that Access is not robust enough for more than about five users, when it has the same database engine as Exchange Server which deals with hundreds or thousands of users (albeit it very badly). "

Same name, same underlying technology, slightly different implementation. I'm running 3000 exchange mailboxes on a dual Piii xeon 733 server, most of which are students, who love using email and the version of Jet which exchange uses doesn't even blink. Nor does it when it's used for Active Directory and I've seen implementations of that which stretch far beyond my modest 3000 users.

One thing to keep in mind is that these are slightly different versions of Jet than what comes with Access, as I said, and also, remember that while thousands of people might be attached to a connector into Exchange via the MAPI interface, LDAP, IMAP, whatever, the amount of different tasks actually accessing the database is a lot fewer; the users don't all talk to the database, they talk to a connector... ok I'm using a lot of poetic licence there... which then goes and talks to the database on the behalf of all it's users in a controlled fashion.

Robert Moir
Friday, November 22, 2002

Fair enough! Anyway if you can run 3000 mailboxes on a dual 733Mhz system, then I can run the 150 on a single processor.

The reason I ask is that I have never been able to get a remotely approximate answer to the question "how many users can comfortably use an Access database?". Now, I know there are a lot of other variables involved, such as the amount of data transferred over the network, and problems with ensuring everybody logs off, so the database can be compacted, but everytime I ask the question I am told go for client server.

Stephen Jones
Friday, November 22, 2002

I've used a lot more than 5 or so people with an access database without seeing problems. I've also seen problems with just 4 people connected. I think that there might be more *potential* for problems with lots of users running access to the same DB over a conventional file sharing setup but that doesn't mean that the potential will always be realised.

Maybe its to do with how many are hammering the database with heavy updates and complex queries, and how many are just trying to run simple queries that read a couple of files out of one or two tables.

Robert Moir
Sunday, November 24, 2002

*  Recent Topics

*  Fog Creek Home