In a comment to my last post, gvwilson asks, "Are you aware that the PATRIOT Act means it’s illegal for companies based in Ontario, BC, most European jurisdictions, and many other countries to use S3 and similar services?"
This is another interesting case of the non-local networked world intersecting with real geography. Not surprisingly, it quickly becomes complex.
I have heard some of the discussion about S3 and the interaction between the U.S. PATRIOT act and the EU and Canadian privacy laws. I’m not a lawyer, but I’ll relate the discussion for other readers who haven’t been tracking it.
Canada and the European Union have privacy laws that lean toward their citizens, and are quite protective of them. In the U.S., where laws are written about privacy at all, they are heavily biased in favor of large data-collecting corporations, such as credit rating agencies. A key provision of the privacy laws in Canada and the EU is that companies cannot transmit private data to any jurisdiction that lacks substantially similar protections. It’s kind of like the "incorporation" clause in the GPL that way.
In the U.S., particularly with respect to the USA PATRIOT act, companies are required to turn over private customer data to a variety of government agencies. In some cases, they are required to do this even without a search warrant or court order. These are pretty much just fishing expeditions; casting a broad net to see if you catch anything. Therefore, the EU/Canadian privacy laws judge that the U.S. does not have substantially similar privacy protections, and companies in those covered nations are barred from exporting, transmitting, or storing customer data in any U.S. location where they might be subject to PATRIOT act search.
(Strictly speaking, this is not just a PATRIOT act problem. It also relates to RICO and a wide variety of other U.S. laws, mostly aimed at tracking down drug dealers by their banking transactions.)
Enter S3. S3 built to be a geographically-replicated distributed storage mechanism! There is no way even to figure out where the individual bits of your data are physically located. Nor is there any way to tell Amazon what legal jurisdictions your data can, or must, reside in. This is a big problem for personal customer data. It’s also a problem that Amazon is aware they must solve. For EC2, they recently introduced Availability Zones that let you define what geographic location your virtual servers will exist in. I would expect to see something similar for S3.
This would also appear to be a problem for EU and Canadian companies using Google’s AppEngine. It does not offer any way to confine data to specific geographies, either.
Does this mean it’s illegal for Canadian companies to use S3? Not in general. Web pages, software downloads, media files… these would all be allowed. Just stay away from the personal data.