Law professor Eric Goldman points to a “really terrible” ruling in a case called National Association of the Deaf v. Netflix that “has potentially ripped open a huge hole in Internet law” by concluding that Internet sites are “places of public accommodation” for the purposes of the Americans with Disabilities Act, simply ignoring a boatload of precedent concluding just the opposite. As Goldman explains, this is apt to unleash a costly and innovation-stifling flood of litigation:

Could YouTube be obligated to close-caption videos on the site? (This case seems to leave that door open.) Could every website using Flash have to redesign their sites for browsers that read the screen? I’m not creative enough to think of all the implications, but I can assure you that ADA plaintiffs’ lawyers will have a long checklist of items worth suing over. Big companies may be able to afford the compliance and litigation costs, but the entry costs for new market participants could easily reach prohibitive levels.

And then there’s linkages with other civil rights statutes, such as Title II of the Civil Rights Act of 1964 (an anti-discrimination law) and state laws, that use similar language as the language interpreted in this ruling. If all of those statutes are back in play too, the range of obligations imposed on websites–and the opportunities for aggressive plaintiffs’ lawyers–expand exponentially. Expect lots of consumer claims that a website discriminated against them based on an impermissible criterion. It’s safe to say that the legal rules at issue in this case could have billions of dollars of impact between the web coding obligations and the potential litigation frenzies.

That wouldn’t be terribly surprising: The ADA has certainly led to its share of costly unintended consequences in the physical realm. But there are a few reasons why the statute seems particularly ill-suited for application to the Web.

At the risk of belaboring the obvious: Web sites are not “places” at all, except in a metaphorical sense: They are streams of information transmitted to users. (There is, of course, some physical place where the server is located, but that place is typically not physically open to the public.) The court treats this as a kind of trivial semantic distinction rendered moot by the advance of technology, even suggesting that it would be “absurd” to exempt Netflix and Amazon from the rules written for the local cineplex and bookshop. But there are actually quite a few crucial distinctions between Internet sites and traditional brick-and-mortar businesses, which make it a rather surprising leap to insist that a statute designed for literal “places” naturally extends to the metaphorical digital “sites” that serve similar functions.

First, information streams, unlike actual places, are intrinsically made of speech. That certainly doesn’t mean the First Amendment is some kind of blanket get-out-of-regulation-free card for businesses engaged in online commerce, but it does mean that expressive rights will often be implicated when changes to the virtual “place” or “service” are demanded, in a way they are typically not when a shop is required to install a wheelchair ramp. Online, the boundary between being compelled to act and being compelled to speak is especially porous. Again, that’s not an absolute bar to regulation, but it seems odd to assume that Congress must have intended to create such broad causes of action against what are essentially publishers of information, when the examples used to define “public accommodation” very conspicuously don’t include any producers of content. “Places of exhibition entertainment” are covered, but not the content of the entertainment itself—which is why the ADA doesn’t require DVD makers to provide caption tracks (though most do) or books and periodicals to run large-print editions. As the court observes, Congress probably could not have anticipated Netflix or Amazon when it passed the ADA in 1990—yet it presumably had at least a passing familiarity with the concept of magazines and videotapes. It does not seem terribly mysterious that Congress nevertheless declined to invite the routine conflict with expressive rights that would predictably result from shoehorning publishers of information under the same regulatory rubric as physical businesses.

Second, physical businesses and their customers—disabled customers in particular—are inherently constrained by distance and the costs of transit in a way virtual “places” are not, which importantly changes the cost-benefit calculus for all concerned. Outside major metropolitan areas, there may be only one or two businesses of a particular type local to some disabled persons, which for practical purposes means they must be able to access that facility if they are to avail themselves of the relevant class of goods or services at all. Conversely, since physical businesses are generally competing only for the patronage of the local population, owners may be more likely to conclude that it doesn’t make economic sense to invest in accessibility aids that will only be needed by a handful of customers.

Things are obviously quite different online. As a general rule, any site on the Web is accessible from anyplace else, which means the inaccessibility of any particular site is less likely to represent a significant practical reduction of the choice set of any individual disabled user. One common argument for imposing the ADA’s universal obligations on physical businesses is that it may be unrealistic to expect the disabled in particular to simply “go somewhere else” if the nearest business can’t accommodate their needs. That argument has less traction in a landscape without distances. But what if none (or very few) of the businesses in a particular sphere implement some particular desired accommodation? Online, that’s much more likely to be evidence that it would be unduly costly or burdensome to do so—and therefore beyond the ADA’s requirements—because sites are typically competing for a national (if not global) pool of hearing and visually impaired customers.

The suit against Netflix is actually an excellent illustration. The National Association of the Deaf is arguing that Netflix runs afoul of the ADA by failing to provide closed-captioning for all of the movies it makes available to “Watch Instantly” online, among other things. Now, there are something like 35 million hearing impaired people in the United States, and more than a million who are functionally deaf. One can imagine the manager of the lone local theater deciding it’s not worth even a relatively small investment to accommodate a handful of disabled potential patrons. It’s much harder to believe that Netflix is just arbitrarily dragging its heels in making its offerings more appealing to such an enormous potential market: To the extent it’s feasible to do at a reasonable cost, they have ample incentive to be doing it without legal prodding—an incentive magnified by the speed with which Web businesses that get complacent fall to more innovative competitors eager to spot those missed opportunities for improvement. (Anyone remember Altavista and Friendster?)

As Netflix’s attorneys point out, the real holdup isn’t, in fact, a lack of will so much as an inability to act unilaterally: They’re obliged to negotiate with the copyright owners of the movies they stream. The court correctly responds that this specific argument doesn’t affect whether Netflix is classed as a “public accommodation”: Rather, it’s a defense they could raise after that determination is made to show they’re complying with the ADA as far as they’re able. But there are reasons to think there would nearly always be such a defense, because the nature of the online economy should generate far stronger pressure for someone to step in and serve disabled users unless it’s infeasible or prohibitively costly—and for incumbent businesses to preempt any new entrants who’d exploit their failure to do so. Congress judged the ADA necessary based on the history of brick-and-mortar businesses where the constraints of space and distance might be thought to attenuate those pressures. Assuming, without further legislative guidance, that the same regulatory regime is appropriate where those constraints are absent is a bit like applying a statute prohibiting animal cruelty in stables to car manufacturers: Cars and horses may serve many of the same social functions, but they differ in the key respect that provided the rationale for regulation. This may be one reason Congress declined to augment its illustrative list of “public accommodations” to include any purely virtual spaces when it extensively amended the ADA in 2008.

Finally, and perhaps most importantly, users are constrained to take physical spaces more or less as they find them. The owner can make changes in the physical environment; the user or patron cannot. If a theater does not provide wheelchair ramps or wheelchair-friendly viewing spaces, there is not a whole lot a disabled patron can realistically do to change that fact at a reasonable cost. Websites are quite different: How they’re experienced is, in practice, always a collaborative fact determined partially by the code served by the site itself, and partially by the end-user’s browser. If you are experiencing this post as a string of written characters in medium-size on a screen, that is not because of any natural fact about the Cato blog as a “place,” but because you have a computer system configured to interpret the binary stream transmitted by Cato’s servers in that way. Your current browser could easily be reconfigured to display it as much larger text; your operating system probably has an option to deliver that text as a computer-generated audio; and with the right browser and peripherals, you could render it as Braille or a series of odors instead. Even assuming a site’s content is rendered in the statistically “normal” way, the fact that this content is delivered to a highly configurable computing environment under the end-user’s ultimate control means it will usually be vastly simpler for adjustments necessary to compensate for special needs to be made client-side rather than server-side.

This is self-evident for at least some cases: Given the widespread availability of browsers and operating systems that can speak written text out loud—however imperfectly—nobody thinks it would make sense to require every individual Web site to provide a spoken audio file corresponding to each page of text. As of 2012, voice recognition software is not nearly good enough to allow automated on-the-fly rendering of movie captions, but that doesn’t mean provision by the provider of the audio-video stream is the only alternative. There are online repositories like the Internet Movie Script Database and (probably more useful in this context) OpenSubtitles.org, providing crowdsourced, time-synched captions for popular movies—combining a client-side solution with a community-based third party resource.

With physical facilities, it will often be obvious in advance which adaptations for disabilities must be provided by the owners of the facility itself (wheelchair ramps; wheelchair-accessible bathroom stalls) and which are more properly the responsibility of the disabled patron and third-party companies (the wheelchair itself). With Web sites, this will not necessarily be nearly as clear in advance, but there’s ample reason to expect that client-side solutions will often be simpler and less costly. Applying the ADA to the Web effectively locks in a presumption in favor of server-side solutions that is highly unlikely to be optimal. Especially given the speed at which Internet technologies and platforms evolve, it seems far more sensible to stand back and let businesses and online communities determine the most appropriate locus for adaptation as new services and platforms arise.