Quidquid latine dictum sit, altum videtur

July 23, 2002

IP everywhere?

A bit more about Rendezvous, which I mentioned in my Macworld reactions. Specifically, check out this interview with Stuart Cheshire which, among other things, has some illuminating ideas about how the future could be.

For example, once setting up a local IP network requires no human intervention, what would stop us from using Ethernet to connect our home theaters? Instead of the massive tangle of cables, why not just digitize the signals and send them as streaming video? (Well, the entertainment industry would complain, just as they did for VCRs, television, radio, and phonographs.) Better yet, now that we can assume reasonably powerful computers on all home electronic systems, we could work up a system where the components talk to each other and the user only needs one remote control. (Contrast this with our current situation, where certain peripherals actually need to bounce infra-red signals off the walls to do things like change the channel on the cable box.)

Mr Cheshire also suggests that one day we won’t need multiple peripheral connection technologies such as USB, FireWire, 802.11b, Bluetooth, Ethernet and the others; just one sort of connector could provide all of that. (Obviously, we’d need to distinguish wired from wireless connections, but wireless connections are easy to set up—no wires.)

That’s certainly do-able from a technical perspective, and it’s worth moving towards for a number of reasons (in fact, one of the stated design goals of FireWire was for it to replace as many existing systems as possible), but I question whether it’s a good idea for things like keyboards and network connections to use the same connectors. If such a thing were to happen, how would your computer tell them apart? Specifically, if something contacts your computer and claims to be your keyboard, how can your computer know it’s telling the truth? The way the zeroconf protocols work, you wouldn’t be in danger from the larger Internet, just the people on your LAN, but in a corporate setting that could be dozens or hundreds of people.

Do we really want to authenticate our keyboards at startup? (Not to mention our monitors and mice.) How would that even work? You’d need another computer to configure your computer to talk to its keyboard.

Of course, that’s all in the future if anywhere. Hopefully, the developers of such a system will do a better job with the security than… well, all the other network developers. #

Deep linking rises again

Joe Clark has written about the implications of “deep linking”, or more specifically the theory that links are “public facts” and therefore cannot constitute copyright infringement. (A theory I supported here last month.)

Mr Clark makes a lot of good points, but I find I disagree with some of his assertions. In general, he seems to be thinking too much in terms of how web browsers—the dominant but not exclusive client software for the web—present HTML documents.

For example, he questions whether a hyperlink can be considered a statement that a given resource exists at a given address, since some web browsers allow embedded scripts to obscure the address from the user, writing: “The recitation of the fact that this document exists at this location can be dependably found only in the HTML source code where the href attribute is visible, not in a browser rendering.” Very true, but that’s not a counter-argument. The fact that software exists which displays web pages in certain ways does not change the essential nature of HTML documents. The statement is being made in the code; the fact that other parts of the code may result in obscuring the statement from users is entirely separate.

Remember, a great deal of web traffic comes from search engine robots which scan the web for pages for their indices. The only common point between their perception of the web and the one presented by web browsers is the HTML code itself.

The robot perspective also muddles the assertion that hyperlinks are functional rather than expressive. Web browsers, for the most part, present a hyperlink as a bit of text that, when clicked on, prompts the browser to load the linked resource. Google, on the other hand, sees the same link as a connection between the containing resource and the referenced resource. It may change the ranking of various pages or add resources to its to-do list as a result of seeing the link, but you can’t claim that those actions are implicit in the link itself. Neither, then, can you claim that the “click to load this page” behavior exhibited by web browsers is implicit in the link. Certainly, concepts of “a distinctive font or color” apply only in web browsers.

On the other hand, one can make a distinction between linking and embedding, which Mr Clark also seems to have missed. If I link to an image (using an “a” or “link” element), it is clearly a separate resource which I am referencing. On the other hand, if I include the image (using an “img” or “object” element), then I am implying that the image is logically part of my document. That could be reasonably construed as infringement, even without mentioning that most web browsers will insert the image into the displayed document. The fact is that the embedding tags imply different things from the linking tags, which even non-browser systems can recognize.

At least, that’s my take on it. Of Mr Clark’s five conclusions, I agree with only the first three. (In particular, number four seems to be the result of some confusion: a URL is not by itself a statement of anything, complicated or otherwise. As for whether hyperlinks are expressive, functional, or both, it really depends on what “functional” means.) #