terriko: (Pi)
In the course of my thesis work, I made myself a little Firefox plugin that tells me where the javascript/dynamic parts are in a page. It's a fun little thing, just puts some big coloured boxes up, and I used it to help understand how people were using javascript in practice. It's one of those things I should probably release just 'cause it's fun, but I didn't have time to maintain in any meaningful way so I didn't get around to it.

Anyhow, I pulled it out last week to see what state it's in because I want to adapt some ideas from it, and it wasn't working. Which is odd, 'cause it's really quite simple. The core is just a loop that goes through each page element and looks for stuff like onmousover events:


var allTags = document.getElementsByTagName("*");
for each (var tag in allTags) {
// ... do some stuff
}


And in debugging it, I've learned that getElementsByTagName("*"), which apparently used to return all the tags as objects, is now returning all the tags as well as, inexplicably, a number. It's not the same number for every page, and most of them seem to be around one thousandish on the simpler pages I was trying to test. Which sort of makes me think that maybe it's returning the number of tags, or that it sometimes returns an ordinal index for a single tag instead of an object, but why?

As it turns out, it didn't take much to get my add-on back up and running, just a quick check to see if the "tag" in question was in fact an object. But I'm left with a question: why has this changed in Firefox since I initially made the add-on? I'm not even sure where to ask, since it doesn't seem like it's a thing that changed in the specs. I'm recording it here for posterity so I remember to try to look it up later, but if you happen to know what's going on, please get in touch!
terriko: (Default)
I'm re-reading Richard Hamming's talk on You and Your Research because I felt like I needed the kick in the pants to do great work this month after some very busy months of doing necessary but not necessarily great things.

In this reading, I was struck by this anecdote:

John Tukey almost always dressed very casually. He would go into an important office and it would take a long time before the other fellow realized that this is a first-class man and he had better listen. For a long time John has had to overcome this kind of hostility. It's wasted effort! I didn't say you should conform; I said ``The appearance of conforming gets you a long way.'' If you chose to assert your ego in any number of ways, ``I am going to do it my way,'' you pay a small steady price throughout the whole of your professional career. And this, over a whole lifetime, adds up to an enormous amount of needless trouble.


On a surface level, I've long believed this is true. I've been long primed in the art of social hacking, first by my father and more recently as a security researcher/hacker. Anyone can watch the subtle variations on how I dress on teaching days or days when I'm going to the bank and you'll note that I pay attention to fitting in to the environment and manipulating the way in which I'm perceived. But as a child of the Internet, more or less, my experimentation hasn't limited to physical presentation. Especially as a teenager, I spent a lot of time grossly mis-representing my age and gender as well and watching how that changed my interactions with folk.

But what gets me this time is the end of that quote: "[If you don't appear to conform,] you pay a small steady price throughout the whole of your professional career. And this, over a whole lifetime, adds up to an enormous amount of needless trouble." Sometimes it's important to change the system, but sometimes you just want to get stuff done.

I can dress the part, but I don't generally change my gender presentation in real life. Is my female-ness adding up to an enormous amount of needless trouble over my lifetime given that I work in a field where that's going to make me non-conforming? I suspect it is, although I'm fortunate enough that my gender presentation is often canceled out by my racial makeup (Asian girls are totally good at math, don'tcha know?) so I can console myself by saying maybe it's not as enormous as it might have been. But not every person who doesn't fit the norm for their field has that consolation prize. Are we all paying the price of being different?

It's easy to get a little saddened by this. All that time explaining that no, I really am a techie, has added up to a lot of time I'm not having amazing conversations and doing great work. But before you get too saddened about how your hard-to-hide features like race/age/gender are affecting your ability to Do Great Things, you should stop and listen to Duy Loan Le's excellent 2010 Grace Hopper Celebration Keynote. In it, she talks about what she does to fit in to environments where she felt that letting go of her ego made it possible for her to get more good work done. I think it's really worth a listen, especially if fitting in isn't just a choice of what suit to wear for you.

terriko: I am a serious academic (Twlight Sparkle looking confused) (Serious Academic)
When I was an undergraduate, I found that university really wasn't living up to my expectations of stimulating, interesting people and ideas.

But today, I was totally living the academic dream.

We had a visit from a leading expert on ant behaviour. This wasn't about computer ant algorithms; she studies real live ants. We started off the day with her talk on the Turtle Ants she's been studying in Mexico, a talk filled with pictures of ants and paths and grad students on ladders pointing at the trees. A talk filled with speculation about behaviour and patterns and analogies to search in computer networks and bifurcation of biological trees. Over the course of the day, the group talked ants, bees, simulations on the computer and using robots, immunology, flu and t-cells in the lung, patterns and theories. It was the kind of conjunction of ideas from multiple disciplines where things were just clicking and questions and potential experiments started getting debated.

Biochemistry from my scientist parents, ecology and field work from Macoun Club, immunology from the above plus my own master's research, algorithms from math and CS... I was pretty proud of myself for knowing the jargon pretty much across the board and being able to keep up. I love that I'm with a group where seemingly disjoint backgrounds are consistently recognized as a huge advantage, and my own particular background fits right in.

I learned a bunch about ants and flu today. My notebook is filled with doodles of ants and cells doing stuff. Apparently turtle ants, since they have paths in the trees, sometimes get the paths broken when the wind blows, and the ants just back up and wait for the wind to blow the branches back so they can keep going. I learned that swine flu's replication rates in cells are a hundred times higher than avian flu (and ~20 times more than regular flu) but avian flu does other things to suppress immune response. I learned some about how T-cells get into the lungs and find infection despite the fact that they don't seem to move fast enough to explain how well we handle infection. And I got to watch people putting ideas together in ways that might result in using experiments in ants to try to explain things that would be much harder to test in the lungs, and so many ideas that probably just couldn't happen anywhere else.

So if you've been wondering why the heck I moved here despite the many downsides about the US/desert/altitude/regional poverty/city, etc.... this is why: Cutting edge research at the conjunction of biology, computing, and maybe a few fields besides. Even if I decide to do something else once my contract is played out, this has already been amazingly worthwhile, and with my own project starting to take shape, I'm pretty sure it's just going to get better!
terriko: I am a serious academic (Twlight Sparkle looking confused) (Serious Academic)
This was originally posted on But Grace, but I don't want my regular blog to wind up devoid of technical content so I'll be crostposting all my posts from there in their entirety, I suspect.

One of the cool things going on at work is some software we have that automates the creation of small bug fixes. We're looking to try it on some more active projects with real bugs, but we need projects with reasonable test case coverage so that the automated system can also ensure that it isn't causing other things to break in making the fix. Basically, we're potentially offering up a bunch of free bugfixes if your open source project has decent test cases. Pretty good deal, I hope.

Open source projects with good test suites



But how do we find software with good test cases? Here's a few I know of off the top of my head:

Open source projects with good test suites


Name/URLTest suite?Notes
Firefox A brief search turns up some automated tests My fuzzy memory suggests there was more than these...
Gnumeric Extensive regression tests for each function The function tests are in .xls spreadsheets, so we could potentially apply them to other spreadsheet software.
SQLite They claim extensive test coverage Very promising!
Webkit (Chrome, Safari) a brief web search turns up regression tests for javascript I believe the Chromium project has even more tests


Can anyone suggest other software or more details (and better links) on the things I have mentioned already?

Open source routing software



For various reasons, I've been encouraged to try experiments on open source routing software. There's some existing academic literature on the types of bugs found in open source routers, and it seems like our automated patch creation system would be a good fit especially since router bugs often cause huge outages or security problems and having a temporary patch to solve the problem right away could be a huge boon.

My query on twitter generated a nice list of open source router software, but no one seems to know anything about test suites. Here's a table summarizing what I've found thus far:

Open source router software test suite information



Name/URLTest suite?Notes
Click Unknown Nothing obvious, and given that it's on a university website, I'll be shocked if it has testing. ;)
dd-WRT Unknown Nothing obvious in the wiki, but there were lots of hits I haven't investigated.
OpenWRT Unknown Clearly there was interest in automated test suites in Jan 2011 but it's unclear to me if these are now around somewhere. Need to look more.
pfSense No evidence of a test suite Searching the dev wiki for "test" yields nothing likely, so I'm guessing there isn't one.
Quagga There is a tests/ directory, but it looks unsuitable "make test" doesn't work and "make check" pokes a bunch of directories but doesn't seem to do what I need. There's a directory called tests/ in the repository, but I'm not sure what it does. I can run the tests manually, but the output is currently meaningless to me. No one answered my question on #quagga, although another open source friend on #kernel.org suggested that the test suite may have been abandoned.
Tomato No evidence of a test suite The web site contains nothing useful, so if there is a test suite, it's likely being provided by someone else.
XORP There is a tests/ directory, unsure if suitable but look promising These look promising, but I'm having some build errors and haven't been able to run them yet


You'd think, perhaps, that reasonable test suites for routers would already exist. A generic test suite would be totally sufficient for my needs at the moment. And in fact, I've found a set of routing tests from the University of New Hampshire InterOperability Laboratory, but while their tests are well-described, it doesn't look like something we can run locally and repeatedly as we'd need to in order to test the auto-generated patches. I haven't yet found others.

Let's be clear: I don't really care what the router supports in great detail. The important thing for these tests are that there be a good test suite, and preferably a good bug queue so we can grab candidate bugs and bug test cases to try to solve them. Generally speaking, the bugs have been easier to find than the regression tests.

In summary...



I am looking for:

1. More information about router test suites.
2. Updates to my current tables of information. This represents a morning's work, so I'd be shocked if they're perfectly correct.
3. Any open source/free software projects with good test suites (and preferably good bug queues).

Again, the key here is that I need good test suites. I'm most interested in routing software at the moment, but I'm building up a list of alternative ideas if that doesn't pan out, so anything with good automated tests I'll be able to run repeatedly is of potential interest. We've got access to a reasonable amount of computing power, so heavier weight tests are fine as long as they aren't going to take all month to run.

We would love to contribute any fixes we find back to the community, so if you think your project might qualify please get in touch! I think the end result is going to be awesome for all involved: free bug fixes for the project, more impressive real-world validation of our automated patch creation system, and maybe even an academic paper out of it for some of the folk around here.
terriko: (Default)
As some of you may know, my last paper was on visual security policy (ViSP), a neat idea I had about how to add security policy to a website in a way that was more in line with how sites are designed. I based it on my own knowledge working as a web designer, as well as ideas from a variety of friends who have or do work in the web space professionally and not.

You can read my presentation the larger run-down or read the paper, but the idea behind ViSP is that it's sometimes very useful to subdivide pages so that, say, your advertisement can't read your password, or that funny video you wanted to embed doesn't get access to post blog entries, or whatever. Sadly, right now anything embedded in the page gets access to anything else unless some awfully fancy work has been done to encapsulate parts of the page. (And given how much people tend to care about security in practice, this doesn't get done as often as it should.) We currently just trust that any includes will play well, which is super awkward since malicious code can be inserted into around 70% of websites and you can't very well expect malicious code to play nicely.

Anyhow, I digress.

I'm updating this particular policy tool so that I can generate some policies to test, because I'm tired of building them manually, and my not-terribly-scientific method of clicking randomly on things to make policy has turned up a problem: what happens if you want to set policy for an element that's just one of many paragraph tags or whatever, not assigned an id?

With ViSP, we assigned an index based on how many such tags we'd seen, but I figured while I'm updating this surely I could find something more standard...

Turns out, no, that really is the best way to do it. At least according to the selectors API, which includes a nth-of-type() pseudo-class that seems to do pretty much what I want.

So now, if you're using my tool and wanted to define policy for a given tag, any given tag, we can make that work for you by building up a CSS selector to find it. Of course, it'd probably be cleaner to read if you only set policy on tags with ids or classes, but I don't have to require that as an additional hurdle to policy creation. I figure this is likely a net usability win when it comes to policy generation, and let me tell you, security policy is not a field known for usability wins. (So much so that if I google search for the words security policy and usability... I see a post by me on the front page suggesting usability studies on CSP.)

Anyhow... Thanks to having to learn querySelector earlier, I was already primed to create querySelectors for uniquely defining tags. Thanks Mozilla documentation! You're a terrific coding wingman, introducing me to all these awesome apis. ;)
Page generated Aug. 21st, 2014 12:04 pm
Powered by Dreamwidth Studios