FTC: Consumer Privacy Comments Concerning Voters Telecommunications Watch--P954807
Final Comments to the
Office of the Secretary
July 11, 1997
Few issues have been as hotly debated on the Internet recently as unsolicited commercial email. Just as the cost of telephone or cable service, the issue of junk email touches everyone with an email address and has brought forth opinions on how to deal with it in a volume unseen since the debate raged over the Communications Decency Act.
This paper is a sampling of the costs and technical methods of dealing with the junk email issue. It describes the various ways in which junk email costs Internet users and Internet Service Providers, and the technical ways that they are dealing with it today. It brings together some disparate data collection efforts underway by EFF-Austin, EF-Florida, and the Voters Telecommunications Watch.
I.B. Survey Overview
In preparation for the Federal Trade Commission's June 10-13 workshop on Consumer Online Privacy, we prepared two surveys to form the basis of our filings. First and foremost, we wanted to gather some basic factual data about the state of the junk email issue. Anecdotal cost, volume, and technology information seemed lacking in the debate. In fact two of the FTC's questions seemed interested in the costs and technological solutions to the issue. It quickly became obvious that there was a dearth of basic knowledge about how much junk email costs individuals. In an issue where emotions and opinions run high, a lack of knowledge can be fatal.
The survey was begun in mid-April, right before Tax Day. At its height, the survey received over 300 responses per day. The survey was designed to take the respondent's answers and send them to the survey coordinator and the respondent in email form. The presence of email filtering technology available in Eudora 3.0(1)
on the survey coordinator's laptop made this survey somewhat self-referential, as the problem of junk email is a subset of a larger problem of how to manage large amounts of incoming mail.
Over 2,700 people answered the user survey, and 60 Internet Service Providers (ISPs) answered the institutional survey. They are truly the heroes of this issue, as every one of them enriched all of our knowledge of the facts of this issue.
I.C. A note on statistical relevance
None of the staff at VTW, EF-Florida, or EFF-Austin working on the survey has the requisite background in statistics to allow us to make any claims about the statistical relevance of this survey. Furthermore, our survey population is not representative of anything, except the cross-section of people that visited our website. Because of this, we followed a "minimal intrusion principle" and did not attempt to collect any demographic information to shape a representative sample of Internet users (whatever that is). In particular, whenever we say "XX% of people in the survey believe X", it should not be assumed that "X% of the Internet community believes X" or "X% of Americans believes X".
Readers should keep this in mind when reading the results of the survey and consider the benefit is in its breadth of coverage, not its ability to predict the opinions of Internet users on the junk email issue.
I.D. A note on privacy of survey respondents
We look forward to the comments from the Center for Democracy and Technology on consumer perceptions of online privacy to tell us more about consumers' attitudes towards taking part in surveys like this. Concerns over the privacy of respondants kept us from creating a more intrusive (and useful) survey.
II.A. Survey Responses
Over 2,700 people answered our user survey. We didn't require answers to every question in the survey. We present the raw responses here for your perusal.
Question 2: How much email do you receive per week? (only one response allowed)
1-50 pieces: 851
Question 2.2: How much of it is unsolicited email? (only one response allowed)
It is a small amount of my email:
Question 3.1: Is it a risk or a benefit? (only one response allowed)
It is a risk: 1711
Question 3.2: Elaborate on the above (anecdotal answers)
Cited the fraudulent nature of
much junk email: 643
Question 4: How do you deal with unsolicited commercial email? (multiple responses allowed)
I read it like anything else: 367
Question 5: What does junk email cost you? (multiple responses allowed)
It costs me nothing: 197
Question 6.1: What do you think of unsolicited commercial email? (one response allowed)
It is an overall good: 47
Question 7 & 8:Technological and regulatory suggestions (anecdotal responses)
Unfortunately, many people typed the wrong answers into each of these sections. Many typed technical suggestions into the regulatory section, and vice versa. The answers have been aggregated here.
Keep the government out of the
III. Costs of unsolicited email
FTC Question #2.18 What costs does unsolicited commercial e-mail impose on consumers or others? Are there available means of avoiding or limiting such costs? If so, what are they? (2)
The survey we have developed addresses the questions of cost in many different forms, both for Internet users and Internet Service Providers. This gave us a broad view of the problem, touching on many individuals and many people within the industry. We look forward however to the comments from Simona Nass of Panix (Public Access Networks) to provide us a close look at the actual costs as assessed by a particular ISP when dealing with unsolicited commercial email.
III.A. Cost to consumers
The costs to consumers of unsolicited commercial email fall into a number of broad categories that we have enumerated in the survey, including but not limited to: telephone toll charges, ISP connect time charges, and ISP download charges. We separate the costs into the following categories:
hard costs: measurable in dollars of resources, access fees, or staff time and
soft costs: measurable by loss of productivity. These can also be thought of as opportunity costs, where the dedication of resources to dealing with junk email remove the ability of the user to pursue some other goal temporarily.
The comments in this section refer to the attached figure, entitled, "Costs of Spam to Internet Users".
III.A.1. Telephone toll charges - hard cost
For some Internet users, online access is not a free, flat rate telephone call. For those users, the proverbial "meter" is ticking when they are downloading, reading, or even deleting without reading, unsolicited commercial email.(3)
III.A.2. ISP Connect Time Charges - hard cost
There has been much discussion in the market about pricing of Internet access for both large (such as America Online) as well as smaller Internet Service Providers, such as Panix. The amount of time a user is connected to the system with some services is metered at a usually low rate, and the time it takes to read and even delete unsolicited commercial email adds costs to the user. The market for Internet access has become mixed enough between flat-rate pricing and metered pricing that we received varied answers for what connect time costs them, with it being anywhere from US$0.50 to US$4 per hour.(4)
III.A.3. ISP Storage Charges - hard cost
Although it is not a major part of the pricing model today, several respondents indicated that they are charged on a per-megabyte basis for email stored at their ISP's site before the user has read it. In all cases, respondents indicated that they had a minimum of un-tariffed storage available to them before charges kicked in. However unsolicited email adds to this cost, forcing the user to pay more for their Internet service. The range of responses stated that this cost is approximately US$1 per megabyte.
III.A.4. Lost productivity - soft cost
By far the most popular soft cost cited by respondents was loss of productivity. For many, this was a major problem as their time is measurable in US$/hour and they can easily do the math to see what each junk email costs them. Depending on the effectiveness of filtering techniques and the mail package employed by the user, junk email could take anywhere from no time to several minutes to deal with. Although not all respondents mentioned it, one should consider the litany of factors required in dealing with junk email that go beyond the simple act of deleting junk email. If the user downloads their email to a remote computer instead of reading it at the ISP, then the additional time required to download the junk email is time that could be better spent working. Also, the time taken to respond and unsubscribe from junk email lists, or to report such activity to the one or more ISPs involved can be significant.
III.B. Costs to Internet Service Providers
Very few, if any, of an ISP's infrastructure costs can be broken down as easily as those for Internet users. Complicating the issue is the fact that the Internet access market is divided into two major pricing models: metered and flat rate pricing. For metered pricing, the issue simply becomes a problem of user annoyance. However in the more popular environment of flatrate pricing, unsolicited email becomes a threat to the cost structure of the ISP.
In a flat rate environment the presence of a significant resource consumer (such as unsolicited mail) that is unwanted becomes an expense that raises the cost of the ISP to run the site, a cost which is likely to be passed onto consumers. How much this added cost adds up to is unclear, as reliable numbers on the volume of unsolicited mail are not available. Exactly how much of it is unsolicited vs. unwanted is also a factor.
Outlined below are the costs of unsolicited email to ISPs, though we hope to learn more from the filings of ISPs who have actually measured the hitrate on their filters and compared it to their total mail volume. The comments in this section refer to the attached figure, entitled, "Costs of Spam to ISPs".
Note that the costs are sometimes incurred even if neither the sender nor the receiver is a customer of the ISP. In the case of a spammer "relaying" their messages through an ISP on the way to another recipient, the ISP being used as a relay hosts neither the sender or the receiver.
III.B.1. Network bandwidth cost (soft cost)
ISPs pay for their Internet connectivity, usually in a "maximum capacity" fashion. Although a particular message (or set of them) will not trigger a cost to the ISP by their next tier provider, the growth of their traffic load may mean they must upgrade their connectivity to adequately handle their load. With some providers the ISP pays by actual usage, with the type of connection dictating the maximum capacity. In such a case, passing spam would actually incur a real cost.
Such costs need to be examined further, and we hope that filings from ISPs will illuminate this fact.
III.B.2. Cost to process mail (soft cost)
Many ISPs have reported that heavy loads of spam have delayed or prevented other, non-spam mail from getting through to their users. Providing several hosts to process mail can certainly alleviate this issue, though that may be expensive and needless if the users don't want the email to begin with. In addition, because the overall amount of mail sent through an ISP is very large compared to the perceived volume of spam, it isn't clear if a need to upgrade such infrastructure might not be required anyway.
Instances of actual outages exist, though, and the loss of reputation and service is a very real and painful thing for any ISP to bear.
III.B.3. Cost of staff time for handling junk email
As many ISPs will tell you, the most expensive element of their operation is the cost of their highly trained staff. Customers that receive unsolicited electronic mail want to report it to someone. In addition the surges of system activity that accompanies large influxes of junk email can cause disks to fill up, and mail to stop working until it is attended to by a staff member.
In addition, maintaining filters for screening spam cost a significant amount of resources, as do the processing and tracking down of those responsible.
III.B.4. Cost of storage
ISPs typically store email until customers pick it up. It is typical customer behavior to avoid deleting anything until the very last minute. For this reason, many ISPs charge their users for disk space over a standard allotted amount. For the same reason, several ISPs have configured their systems to deliver customer email into the system directly into user-billed storage, which means that you potentially might be billed for electronic mail that you have not yet read.
III.B.5. Cost of phone line availability
On systems where customers are encouraged to dialup quickly, download their mail and then disconnect, large amounts of junk email cause users to stay on the line for longer, tying up dialup lines for longer. Although ISPs currently do not pay per-minute costs on incoming calls from subscribers, they do need to ensure they have enough telephone lines to support their users.
IV. Responses to unsolicited email
FTC Question #2.19 Are there technological developments that might serve the interests of consumers who prefer not to receive unsolicited commercial e-mail? If so, please describe. (5)
IV.A. Methods of dealing with unsolicited commercial email
At this time, the technological innovations for dealing with unsolicited email are in their early stages. We can see from experience that no single solution (litigation, regulatory, or technical) can adequately resolve the issue to the satisfaction of the Internet community. However what isn't clear is whether or not an appropriate combination of all of these solutions will be sufficient.
IV.A.1. Reading it or deleting it
It is undeniable that a certain segment of the Internet community actually reads or simply deletes their junk email with a minimum of effort. What is currently being overlooked in the debate over unsolicited email is whether the objections to the practice are because of the costs such mail brings, or because of the often fraudulent nature of such mail.
IV.A.2. Filtering in theory
The unsolicited bulk email issue may partly be a subset of the problem that most people receive far more information than they can handle. The use of filtering to get a handle on dealing with large amounts of mail predates the problem of junk email. This is evidenced by the fact that so many products contained filtering technology long before unsolicited email became a front page issue.
Filtering is done at any number of points in the delivery stream, from the end user's mailreader all the way up to the router that connects the ISP to their upstream provider. However filtering at each point is a tradeoff between granularity and cost savings. For example, were an ISP to throw away all mail from spammer.com at their upstream router, the entire customer base would not have to deal with any unsolicited mail from spammer.com. However because there is no granularity at this point, individual customers could not "opt in" to receive any mail (unsolicited or otherwise) from spammer.com. Were the filtering done farther downstream, for example during the local delivery process, the user would have the choice of opting in or out of it. If every user were to choose to have spammer.com mail not sent to them, the effect for the end users would be the same.
However the ISP wouldn't see the savings in terms of bandwidth or CPU, since the mail would still need to be delivered up until the last minute, when the users' preferences would be taken into account. They would see the savings in terms of disk space and connect usage, as the mail would never hit the user's end mailbox.
The final place to filter is in the mailreader. Many mailreaders (see "Mail filtering techniques" table) have the ability to filter electronic mail in many points of the delivery chain.
Regardless of where filtering is done, it tends to happen in one of two ways: opt-in filtering or opt-out filtering. We have found that most filtering mechanisms allow users to implement both kinds of filtering without difficulty, although varying degrees of technical knowledge are required. The comments in this section refer to the attached figure, entitled, "Spam filtering points".
IV.A.2.1. Opt-in filtering
Opt-in filtering is probably not the most popular one used, but it is the most effective at limiting the amount of email one deals with. A user by default receives no mail unless they "opt-in" by specifying that particular email that matches a certain criteria is to be delivered to the user. Email that does not match the criteria is deleted or left in a "junk" folder to be read later. This addresses one of the risks of junk email that was widely cited in our survey: that junk email will clog up one's mailbox, making it harder to find truly important messages among the rest of the mail.
IV.A.2.2. Opt-out filtering
Opt-out filtering works by assuming that the user wants to receive all mail, and then selectively deletes or adjusts the priority of mail that matches particular criteria. Although opt-out filtering is less likely to hide or delete messages that you might really want to see, it requires ongoing maintenance to keep the opt-out criteria fresh.
IV.A.2.3. Filtering granularity vs cost
There is an interesting zero sum game played out when one looks at where to do filtering. Referring to the "Spam filtering points" diagram, one can see that choosing where to filter will be a balance between costs and granularity, even though in the end, the user will not see the junk email in their mailbox. At the far left end of the continuum, one can make filtering decisions only for the entire network which will results in the most savings to both users and the ISP. However the decision to filter out a particular email source will be for the entire customer base. If some small amount of customers wish to receive email from the source, they are out of luck.
On the other hand, if the filtering is pushed to the far end of the delivery process, when the email hits the end user's Inbox, then the maximum amount of resources are spent during the delivery of the mail, but each user has the maximum amount of choice regarding what mail they do and do not receive. Decisions made by each user do not affect what any other user is allowed to read.
This tension between the costs of the ISP and the convenience of their customers is likely to get much worse before it gets better, with large making blanket filtering decisions for the entire user base at the perimeter of their network. In some cases, these decisions will surely be inappropriate and result in criticism of the service by its customers. How these decisions are handled will certainly enhance or diminish the image of the provider in a marketplace as diverse as the ISP marketplace.
IV.A.3. Filtering in practice
In practice, filtering is far messier than the theory would suggest. Given the Internet as a static medium, filtering would be easy, but nothing is true for any length of time, neither the domains that send you junk email, nor the responses people use when they want to show their anger about it. Both the architecture and effectiveness of filtering schemes change on a daily basis, with the same scheme being highly effective one day, far less effective the next, and then effective again on a third day.
IV.A.3.1. Filtering at the network connection
This would typically be done through a router at the perimeter of the ISP's network. Because there are more addresses that send non-junk email than there are junk email domains, this is typically done in an opt-out fashion. When enabled, this configuration disallows all network mail service connections from a particular set of Internet addresses, regardless of who the intended recipient is. Decisions made at this point cannot take individual user preferences into account.
The most likely network connectivity filtering we'll see in the short term is severing of "peering" agreements between ISPs, when one ISP begins to support a customer base of unsolicited commercial emailers.(6)
IV.A.3.2. Filtering at the mail receipt level
This consists of configuring one's mail software to refuse mail from selected sources, regardless of who the mail is for. Like network level filtering, it is also done as an "opt out" system because of the problem of knowing a priori the source of non-junk email messages. Mail receipt filtering has the advantage of allowing a large network to implement different system-wide mail policies for each system within the network. Individual user preferences are not taken into account at this point.
IV.A.3.3. Filtering during delivery
This is one of the most popular points to do filtering, as it allows the mail to be discarded during the delivery to the individual user's mailbox. Using tools such as "procmail"(7) , each user can set their preferences or even delegate their decision-making power to the ISP for individual choices. If the user knows the domains from which he or she will be receiving mail a priori, they can set their filters up in an opt-in fashion, though this is certainly not the way most people have done it.
A user can even choose not to filter at all, even if the rest of the customers choose to.
IV.A.3.4. Filtering right before or after end user delivery
This is the traditional place where filtering is done and is built into several different types of mailreaders, including the highly popular Qualcomm Eudora and Microsoft Exchange. As with delivery time filtering, it can be done as either "opt-in" or "opt-out". Opt-in filtering can work quite well at this point, because it is technically feasible to segregate messages from known-safe sources into a special "Personal" folder, and everything else into a lower priority folder to be read later, instead of deleting it. This allows the user to truly make a large amount of mail manageable.
Filters take effect as the mail is downloaded (Exchange's Inbox Assistant) or right after (Eudora's Mail Filters).
IV.A.4. Specific filtering technologies
We list below several examples of technologies that can perform filtering at various levels for the ISP or the user. We do not endorse or condemn any particular product for use in filtering junk email.
IV.A.4.1. Filtering technologies at the network level
Most network routers have the ability to reject traffic from a specified list of Internet sites, though some do it faster or more efficiently than others. The major variances among these techniques involve the source of the Internet addresses to filter out.
The most effective way of defeating this filtering technique is for the spammer change their IP addresses (or ISP) on a regular basis.
IV.A.4.2. Specific mail receipt filtering technologies
Unix 'Sendmail'(8) is one of the most popular perimeter mail transfer agents in use today. It has pre-built commands to allow an administrator to reject mail from specific domains.
The most effective way to defeat this technique is to either lie about the domain that the mail is coming from, or to register a brand new one every time one needs to send junk email.
IV.A.4.3. Specific mail delivery filtering technologies
Filtering technologies at the delivery point are one of the most effective and form a "sweet spot" of minimizing cost and maximizing user preference. Not only can users configure their own filters, but because all users tend to share a common filespace, they can "share filters" with each other or even the ISP staff.
Through programs like "procmail" and AOL's "Preferred Mail"(9) , a user can choose to have filtering done, but isn't required to maintain the filter list themselves. The ISP has an appropriate motivation to maintain it, since it cuts down on user complaints and all the costs that come later in the message delivery stream after delivery.
The most effective way to defeat this technique is the same as the receipt level: falsify the Internet source domain or register new "throwaway domains" every time one needs to send mail.
IV.A.4.4. Specific end-user delivery filtering technologies
This is the most popular mail filtering technique. Programs like Eudora and Exchange have user-configurable filters that allow any user to setup fairly complex rules that can segregate mail into different mailboxes, set priorities on messages, and automatically take actions on them (including deleting or reforwarding the message to others). The only drawback of performing filtering at this stage is the ongoing obligation to maintain the filters.
IV.A.5. Response tactics
Several respondents reported that they take measures to reach the senders of unsolicited electronic mail or their Internet Service Providers. These tactics vary from the effective to the dishonorable. They are not necessarily technical in nature, but are presented here for completeness.
IV.A.5.1. Contacting the ISP of the unsolicited email sender
One of the most effective non-technical methods for ending unsolicited commercial email involves reporting the spammer to their Internet Service Provider. In our ISP survey, almost all (51) of the ISP respondents said they had a policy against the use of their services for the sending of bulk unsolicited email, and over half (27) said they had terminated a customer because of it since the beginning of 1997. Many ISPs maintain a specific alias (usually "abuse") for individuals on the Internet to report problems. Several ISP respondents cited as many as eight distinct incidents of junk email that have had to be handled since January of 1997.
We collected some anecdotal data on the cost of handling these incidents and found that cost of staff time per incident is measured in hours, usually 4-7. Respondents provided wide variations in the dollar cost of each incident, though many stated an hourly cost of staff time at about $120 / hour.
Because of the high cost of handling such incidents, ISPs do not seem to hesitate to terminate their contracts with users who send unsolicited bulk email through their services. Several respondents reported that they took great pleasure out of reporting a spammer to an ISP and receiving a notice that the spammer's account had been terminated.
The new practice of spammers starting their own ISPs has particular policy implications, which are not discussed in this paper.
IV.A.5.2. Electronically attacking the sender or their ISP
It has become an unfortunately common practice to electronically attack the sender or intervening ISPs to "punish" them for unsolicited bulk email.
Without judging this particular tactic, it is important to note that:
because of forged domains, the wrong person sometimes gets targeted for revenge, we suspect that this sort of activity is illegal, and if even 1% of the recipients of junk email retaliate with 100 such letters of retaliation, the retaliatory response becomes just as big a problem from the ISPs point of view as the original spam.
V. EFF-Austin Spam Analysis by David Smith
Since July 9, 1996, EFF-Austin's I decided that instead of deleting spam from my personal in box, I would start saving the unwanted e-mail to a separate folder. There was no particular reason for deciding to do so, other than the thought that this information might be useful some day.
From July 9, 1996 through May 31, 1997, I have personally received 782 pieces of spam.
# of Spams Received
Projecting this data onto a simple linear growth model, I am on pace to receive 1753 spams in 1997.
Spams By Volume
Nearly all of the spam I receive is an ASCII (or sometimes HTML) format though I recently received my first spam containing a multimedia attachment which the sender asked me to run as a stand-alone program.
If this practice becomes more prevalent the kilobytes per month ratio should skyrocket dramatically.
Types of Spam Involved
I selected four months August, October, January, May for a content analysis of the spam I was receiving. I coded the spam into four main categories :
Opportunities -- e-mail whose intent was to solicit my
participation in a money-making venture. Products, Goods,
and Services e-mail whose intent was to sell me
Adult-Oriented Sites, Pyramid Schemes, and Other Fraudulent E-mails
I further extracted two more sub-categories, Adult-Oriented Sites (from Products, Goods, or Services), and Pyramid Schemes (from the category Business Opportunity).
In October I discovered the National Fraud Information Center at http://www.fraud.org.Through examining materials at both their website and the Federal Trade Commission's, I was able to learn the difference between a pyramid scheme and a legitimate multi-level marketing plan as well as how to identify credit repair frauds. Since October I have forwarded approximately 35 e-mails to the NFIC as being either a pyramid scheme or a credit repair kit fraud.
Those are simply the ones that are easy to spot on the surface as being fraudulent. I believe that most of the "Business Opportunity" spams make claims of unrealistic income expectations. Many of the health-related products make unsubstantiated if not outrageous claims about their benefits as well. I have also received solicitations for financial services that seem too good to be true. A bank in Arizona, for example, run out of a P.O. Box offering certificates of deposit ranging from 18 to 30% interest. While there is no way to fully identify which are fraudulent and which are legitimate without extensive investigation, I would estimate that somewhere between 15 and 20% of the spam I receive is suspect.
Although this paper takes no intentional positions on the appropriate responses to spam, it does seem to lead to some conclusions that must be taken into account. In particular, this survey demonstrates that the costs of junk email are borne by the receiver. This report shows that they may be costly for some end users today, but could become very costly in the future should the volume of junk email rise as shown by EFF-Austin's analysis from the past year. This cost structure is similar to that of the telephone facsimile paradigm.
However, this paper also demonstrates that the Internet as a medium is distinctly different from the telephone facsimile medium, in that there is the capability for individual users and system operators to inexpensively express preferences and subsequently filter content, alleviating some, if not all, of the costs associated with junk email. However there are costs associated with filtering itself, and the presence of the ability to filter does not automatically mean that this will always the best solution.
It is also important to note, without prejudice, that the last time a telephone medium regulation was applied to the Internet over widespread public concern was the Communications Decency Act. We leave the merits of this approach to another paper.
We applaud the Federal Trade Commission for examining this issue in detail, for allowing us to provide a forum for these comments, and for their continuing efforts in the area of consumer online privacy.
VII. About the authors
This report is a project of the Voters Telecommunications Watch (Shabbir J. Safdar), EFF-Austin (David Smith), and Scott Brower (EF-Florida). Indispensable help was provided by Simona Nass and Victoria Fike (Public Access Networks), Professor David Sorkin (John Marshall Law School), Kieran Ringgenberg (VTW), Deirdre Mulligan and Bob Palacios (Center for Democracy and Technology).
Our perpetual thanks go to Martha Landesberg of the Federal Trade Commission for her encouragement and the staff at Panix Public Access Networks for continuing to provide us the Internet facilities through which we did this work. Finally, we owe a great debt of thanks to the nearly 2,800 Internet users and Service Providers who answered our survey. We could not have done it without them.
3. Citing the cost to consumers in remote area like Alaska where Internet access is not a local call, Senator Frank Murkowski (D-AK) has introduced the Unsolicited Commercial Email Choice Act (http://www.senate.gov/~murkowski/commercialemail/ ) which forces labelling of UCE.
5. See FTC Notice Requesting Public Comment and Announcing Public Workshop above.
6. When two ISPs setup a "peering" agreement, they are agreeing to carry each other's customer traffic on their own networks.
7. Procmail is a mail filtering program that allows the user to segregate and screen messages based upon virtually any criteria found in the message. It is highly complex and not recommended for novice users. For more information, see http://www.ling.helsinki.fi/~reriksso/procmail/mini-faq.html
9. AOL's Preferred Mail software filters mail from domains that AOL says are dedicated to sending unsolicited commercial email. For more information and a list of those domains, see http://www.idot.aol.com/preferredmail/