Web service EULAs

THERE APPEARS TO BE a very real danger that the Web services world is evolving into one in which products and services will not be subject to public criticism. And the funny thing is Microsoft seems to be both an instigator and a victim of this threat to free speech.

For some months readers have been spotting the same provision in various Windows EULAs (end-user license agreements), most recently in the license that pops up when one downloads Service Pack 3 for Windows 2000. “You may not disclose the results of any benchmark test of the .Net Framework component of the OS Components to any third party without Microsoft’s prior written approval,” the Microsoft EULAs say.

This provision caused considerable consternation among the readers who spotted it. Some were worried this meant Microsoft plans to install .Net on their hard drives whether they want it or not (not a totally irrational fear since Microsoft is also now claiming the right in the SP3 EULAs to automatically download OS updates — for more on this subject see Window Manager ). All expressed fears about what the benchmark nondisclosure would mean for their right to criticize Web services. “Does this mean saying ‘.Net is as slow as molasses on my box’ is a violation of the EULA?” pondered one reader.

A good question, because it points out just how far-reaching a term like this theoretically could be if .Net becomes everything Microsoft hopes. Not only would you need Microsoft’s blessing to make quantitative comments about .Net performance, you would also have to run it by Redmond if you wanted to use .Net benchmarks to measure performance of computers, or search engines, or whatever. And can any criticism of performance be considered the results of a benchmark?

You may think such concerns unduly alarmist, given that we know these censorship clauses in shrinkwrap licenses have no real legal standing (for more background on the censorship clauses, see ” A censorship test case “). But although such terms in a shrinkwrap license are unenforceable in court, they still can have a chilling effect. The most flagrant example was Microsoft’s use last year of the benchmark nondisclosure term in the SQL Server license to intimidate a small, independent lab into not going public with its results comparing two versions of Windows. But the censorship clauses inhibit open discussion in more subtle ways, as amply demonstrated by the vendor-sponsored Web service benchmarks published in recent months.

In the past year, Microsoft, Oracle, IBM, Sun, and others have been involved in a rather silly benchmarking war pitting .Net versus J2EE and allied database and application server products. First one vendor publishes data on its Web site purporting to show how its current implementation kicks the other guy’s butt, then the other side retaliates with benchmarks of its own. But since all the licenses for the server software products come with their own benchmark nondisclosure terms, each party has a built-in excuse for knocking down a straw man rather than running as fair a test as possible. The upshot is that all the benchmark numbers lack credibility.

Frankly, in an area where much remains in the realm of promises and specifications rather than real products, I suspect attempts to make definitive performance comparisons of Web services development models are very premature. And I don’t know if InfoWorld or any other publication can come riding to the rescue any time soon with benchmark tests of its own. It’s not because the publications are afraid of the censorship clause (any publication worthy of the name would welcome the chance to challenge them). The unfortunate fact is that testing in a cutting-edge area like this can’t be done fairly without each vendor’s cooperation.

I spoke to Microsoft officials involved in the .Net benchmarking efforts and it’s clear they are actually quite frustrated by the situation. They seem sincere in believing that credible benchmarks of current implementations would tell a good story for them. That may well be true (Microsoft has some obvious advantages as the one-stop supplier of all the .Net components), and I told them that I think Microsoft is shooting itself in the foot by bringing the censorship clauses of the database server world into the mainstream of Web services. Their response is that they don’t really like the benchmark nondisclosure clause but they had to apply it to Windows 2000 to protect when the other application server vendors adopted censorship clauses. In other words, they need to “prevent disreputable competitors from fabricating bogus tests” as one Microsoft official put it last year.

Yet, as is so amply demonstrated by the .Net vs. J2EE benchmark wars, censorship clauses do nothing of the kind. In fact, they make it easier for competitors to stack the deck in their favor and then claim they couldn’t do otherwise because of competitors’ license restrictions. A competitor that does publish bogus test results will be far more vulnerable to defamation or false advertising charges than it will be to violation of an unenforceable clickwrap license term.

I’ll remind readers of the notorious SoftRam 95, a product that was revealed to be basically nonfunctional only after a competitor went public with the benchmark results to prove it. Open public discourse about products simply isn’t possible when anyone, even competitors, is expected to ask permission to speak. If Microsoft really wants the truth about Web services architectures to come out, its first step should be to eliminate these censorship clauses from all its license agreements.

Source: www.infoworld.com