Why Gemini, rather than SafeWeb?

Solderpunk has written an essay comparing Gemini with a potential subset of the WWW, termed the "SafeWeb".

SafeWeb is basically what the web would be like if all but nine HTML tags and one HTTP header magically disappeared from all existing webpages and webservers. SafeWeb and Gemini would look and feel pretty similar. The crucial difference lies in what would happen if the magic wore off and the the other tags and headers REappeared: SafeWeb browsers would change back to behaving like WWW browsers today, but Gemini clients' behaviour would remain unchanged.

Let's take the difference between the full WWW as we know it today, and the subset known as SafeWeb, and call it the "delta". This is all the tags and headers of today's WWW, which enable the use of cookies, Javascript, CSS and all the facilities which depend upon them and make up today's rich WWW user experience. The benefits of the delta are obvious, particularly in terms of visual appeal. The tradeoffs are now also fairly well-known: interstitials, surveillance, distracting advertisements, high resource consumption, security risks and so on. Using a system without the "delta" doesn't eliminate the possibility of, say, surveillance by Google, but it does mitigate it by rendering it orders of magnitude more difficult.

Thus there is a trade-off concerning the delta. When we want the benefits of the delta, we can just use the WWW as it exists today; when we choose NOT to have the benefits and costs of the delta, then we are left with Gemini-like systems and with the SafeWeb.

Let us suppose that some of the drawbacks of the delta are severe enough that we want to be reasonably sure that we do NOT have the functionality of the delta. We can close our WWW browser and open up a Gemini client, as Gemini provides an appropriate guarantee that it is delta-free. The SafeWeb, on the other hand, cannot reliably guarantee that the costs of the delta are absent: the WWW browser software affords all the necessary functionality, and the engineering effort to ensure that this latent functionality is unused is much greater than the engineering effort to construct Gemini or a Gemini-like system.

To put it more simply: it is easier to construct a system which lacks the delta than it is to take a system which has got the delta already implemented and ensure that the delta is never used.

Solderpunk's essay on subsetting the WWW