Is there any update on this project? I note the dates are 2007 but a google for the name didn’t yield anything.

Unfortunately, this project has been cancelled by my hierarchy. So much for burning ships…

I have been thinking about similar things, and having trouble grappling with a way to make decent assertions with the dynamic nature of the language.

Well, of course, you’re not going to get away with the full dynamism of the language if you want decent assertions. If you’re looking for bibliography, you should take a look at JS0. You could start from this and progressively expand.

I have been thinking about similar things, and having trouble grappling with a way to make decent assertions with the dynamic nature of the language.

Being able to perform general-prupose analysis on this type of code is only so useful… but being able to add specific rules for JS Classes which are unique to our environment means that we can go further, and potentially perform tasks such as enforcement of business rules, custom and non-trivial security enhancements, etc

It will definitely be possible to add specific rules. For now, however, I can’t think of any method easier to do that than writing a little OCaml code. If you have any example of a specific rule, though, I’ll be glad to read it.

A few static tools for JS exist, with some even in the research (Yu et al. at POPL ‘07, e.g.)

I’ve taken a second look at that paper and it seems I had missed just about everything that was interesting in it. It looks like something we could implement with JStify once we have rewriting tools.

This is indeed an interesting tool. You must also remember, there are applications for it outside of the browser, and indeed outside of the OSS world which make its extensibility particularly interesting.

For example, we use JavaScript as glue or application code driving large collections of well-worn systems code. Being able to perform general-prupose analysis on this type of code is only so useful… but being able to add specific rules for JS Classes which are unique to our environment means that we can go further, and potentially perform tasks such as enforcement of business rules, custom and non-trivial security enhancements, etc.

A few static tools for JS exist, with some even in the research (Yu et al. at POPL ‘07, e.g.), but this promises to be the most useful and general of them all.

Thanks for the reference, I’ll try and read it tomorrow. At first glance, the paper you mention seems to indicate that the Flint team concentrates on a subset of JS and attempts to mathematically build provably sound policies, while we try to go the other way round, start from the full language, with initially probably-somewhat-unsound policies and experiment our way towards sound policies. Both approaches have their interest and I often belong to that let’s-build-mathematically school of thought. It just happens so that I believe it’s not adapted to JavaScript, but I might be wrong.

What are you going to do about the inevitable deviations from the specification? It might be possible to parameterize the analysis framework so that it can be run with an eye to different browsers/platforms.

Good question and good suggestion. I’ll keep both somewhere safe until time comes

What are you going to do about the inevitable deviations from the specification? It might be possible to parameterize the analysis framework so that it can be run with an eye to different browsers/platforms.