Efficiency? What Efficiency?
The 5th Circuit makes an interesting observation in its Central Hudson analysis in footnote 24 of its recent White Buffalo Ventures decision:
…courts’ tendency to favor the server efficiency interest in name but the user efficiency interest in substance.
The 5th Circuit had, properly, rejected the University of Texas’ argument that spam from White Buffalo Ventures impacted server efficiency. The rejection is not proper because UT’s argument was not true, but because it was not proved in the summary judgment record. In order to understand why it’s instructive to look at a couple of well-known criminal cases: State of California v. Orenthal James Simpson and State of California v. Michael Jackson.
Most everyone I know says that they know that OJ Simpson killed his former wife, Nicole Brown Simpson, and her friend. But, OJ walked. Why? The answer is simple if you remember the trial: “If it doesn’t fit, you must acquit.” There was a reasonable doubt based upon the evidence. The prosecution failed to prove its case. And so OJ is trying to find the real killer by stalking the golf courses of Florida today.
The same is true with Michael Jackson. Most of the people I know really believe that Micheal Jackson has acted inappropriately with underaged boys. But, Michael Jackson walked. Why? Again the answer is simple. Even though the state put on some evidence that Jackson had done something with someone, there remained a reasonable doubt as to the truthfulness and motives of the witnesses which translated into a reasonable doubt as to Jackson’s guilt. Thus, people think that he did something to someone that he should not have done, but the prosecution failed to prove that he did that something with this someone.
The question that the 5th Circuit had to answer in White Buffalo Ventures was this: “Does the First Amendment bar UT, as an agency of the State of Texas, from blocking White Buffalo’s advertisements, even if 8(c) allows UT, as a provider of Internet services, to block White Buffalo’s advertisements.”
Once you understand that this is the question the Court is ultimately answering the layout of this decision makes a lot more sense.
Why? Because the first logical point to settle is the second one. If the statutory basis for UT’s block does not stand then the First Amendment question is never reached. Thus, the first part of the decision (after procedural history) is the conflict between 15 USC 7707(b) and 15 USC 7707(c). The first says that this law preempts state laws, rules, and regulations. The second says that ISPs may block with impunity. Since UT is both a State Agency and an ISP there is an obvious conflict here. If 7707(b) controls then it doesn’t matter what the First Amendment says as UT can’t block the email anyway as the statute says it has to accept the email.
But, the 5th Circuit found that 7707(c) controls. This means that the statute allows UT to block. That only leaves the first half of the question to be answered: Does the First Amendment prevent UT — the State Agency — from blocking White Buffalo’s ads?
Thus begins the Central Hudson test. Central Hudson is one of the most commonly quoted cases when it comes to the restriction of commercial speech under the First Amendment. It involves a four part test:
- Is the speech unlawful or misleading?
- Is the government’s expressed interest substantial?
- Does the state’s action directly promotes that interest?
- Is the state’s action more extensive than necessary to promote that interest?
In its decision, the 5th Circuit rushes through the first three questions to get to the fourth although it notes that the question of user efficiency and server efficiency are not the same in its perfunctory discussion of the second prong of the test:
For purposes of evaluating the summary judgment, we acknowledge as substantial the government’s gatekeeping interest in protecting users of its email network from the hassle associated with unwanted spam. Also substantial is the “server efficiency” interest, but it must independently satisfy a “goodness of fit” inquiry under the fourth prong of Central Hudson. “Suffer the servers” is among the most chronically over-used and under-substantiated interests asserted by parties (both government and private ones) involved in Internet litigation, and rules imposed pursuant to such interests require more than a judicial rubber-stamp, for reasons we explain in part III.B.4.b, infra. (pg. 9, emphasis added)
Please notice that the problem is not that “suffer the servers” is used so often or that it’s a bad argument. The problem that the Court identifies is that it is “under-substantiated” — that it is alleged but the allegation is not backed up by evidence.
When it comes to user efficiency UT put on the testimony of Daniel Updegrove (VP of Information Technology). He linked the problems facing users:
- A filter has to be configured
- There is a danger wanted mail being filtered (false positives)
- Taking the time to delete all of the messages one by one
- How much of an email has to be looked at before the determination “this is spam” is made
The 5th Circuit found this persuasive when it comes to user efficiency. That is, administration staff and end users are both more efficient when spam is blocked using blocking lists.
However, user efficiency is not the same as server efficiency. UT tried to strengthen its point by using both. Therefore, both must be proved using sufficient evidence.
As the Court continues its analysis it gets to footnote 24, which is pretty interesting. It is here that the statement we quote at the beginning of this post is made:
…courts’ tendency to favor the server efficiency interest in name but the user efficiency interest in substance.
The substance of footnote 24 is a cautionary note that what is alleged must be what is proved and that courts may not merely act as a rubber stamp for allegations.
This should serve as a practice note to attorneys doing Internet abuse litigation: Make sure you know what you’re proving with your summary judgment evidence. UT’s problem here was that it fell into a trap that a lot of trial courts have fallen into: It proved user efficiency and alleged server efficiency. Trial courts make a perfunctory reference to CompuServe v. Cyberpromotions and it’s (true) statement that:
To the extent that defendants’ multitudinous electronic mailings demand the disk space and drain the processing power of plaintiff’s computer equipment, those resources are not available to serve CompuServe subscribers. Therefore, the value of that equipment to CompuServe is diminished even though it is not physically damaged by defendants’ conduct.
However, the analysis cannot stop there. The same can be said of legitimate mailing requested by the end user. It also ties up disk space and drains CPU cycles. The difference is one of permission. If someone wants to allege server degredation then server degredation needs to be shown in the record.
This wasn’t a problem for UT as it had also alleged user efficiency, but had it only alleged server efficiency they would have lost. When you allege that 2 + 2 = 4, it is important that you prove that. It doesn’t matter much that you did a good, persuasive job of proving that 1 + 2 = 3 because your allegation is 2 + 2 = 4.
So, how would UT go about proving server efficiency? What would it take to convince the 5th Circuit that allowing all and sundry to mail using time and rate limits would still serve to decrease server efficiency? A study.
UT would need to do a study showing the impact on CPU load, memory load, and disk space of implimenting time and rate limits. Some mail transfer agents allow for such customization. Postfix, for example, allows the use of policy servers. A policy server could be devised that would handle rate limits. Another policy server would handle time limitations for IPs wishing to send mail under an agreement with UT.
But policy servers take some power away from the server. For example, I use an SPF policy server written in perl to provide some information about senders. While it runs it uses around 6% of my CPU and .8% of my server’s memory capacity and has a file length of less than a megabyte. SpamAssassin, on the other hand, uses 11% of my CPU and 6% of the server’s memory and is several megabytes in size, including rulesets. If UT were to create policy servers to handle rate and time limits so that White Buffalo could send its email, there would be some server inefficiency created as the policy server(s) would require disk space and take up memory and CPU cycles. This is even more true if a new policy server had to be set up for each exception.
The result of this study would be the gathering of some hard numbers which would show that even if it were “easy” to impliment time and rate limits that the server inefficiency introduced would make implimentation infeasible. This would satisfy the 5th Circuit’s demand.
It is perhaps made more important by the fact that most of the justices on the 5th Circuit have a view of computers which says that they are “black boxes” that “just work”. The task of the movant (UT in this case) is to show the justices that their servers are not things that “just work” but things that require a certain number of resources before they can work efficiently, if at all, and that conforming to every request to “work things out” in some form of “side deal” steals those resources away to the point of inefficiency.
We’ll see if the next group to face White Buffalo’s argument will figure this out.
MickC @ August 8, 2005