Will HTML 5 Promote Insecure Programming? Maybe not.

[Notice for our international readers]

A few days ago the W3C published the first draft of HTML 5. One of the many new features struck me as a possible amplifier for insecure programming: HTML 5 extends the type attribute of the input element to support URLs, e-mail addresses, date, time, and other types. The rationale for the new types reads (emphasis by me):

»The idea of these new types is that the user agent can provide the user interface, such as a calendar date picker or integration with the user’s address book and submit a defined format to the server. It gives the user a better experience as his input is checked before sending it to the server meaning there is less time to wait for feedback.«

Now this is a really old theme in Web (in)security. The Web as a platform for programming invites errors in input validation and sanitation by giving the programmer equally powerful tools for two different domains of trust, the client and the server. Furthermore, client-side input validation does make sense and is desirable under usability considerations but cannot replace server-side enforcement.

Consequently, one all too common mistake in Web application programming is to validate or sanitize data on the client side but not on the server side where one must not rely on any assumptions regarding client behavior. At the first glance abovementioned extensions seem to provoke even more of these mistakes by improving on the client-side features, thus making them more attractive.

The new feature makes generating code easier, though, which means it may become easier to develop and use frameworks instead of hand-coding. This would be good, security-wise, as one framework usually makes fewer errors than hundreds or thousands of programmers.

At this time, both theories seem equally plausible to me. Empirical studies, anyone?