The entire US Code is now online in XML (opens in new tab)

Likewise in the UK:

http://www.legislation.gov.uk/developer/formats/

XML, HTML, RDF/XML for everything, as well as browsable online (often with original source PDFs of the printed laws), with both online and RDF/XML representations showing all alterations to the law (with date, cross-reference to the Act that made the amendment, etc.).

The website is one of the big success stories of RDF, IMO, as it is all based around a model of the laws in RDF, with everything else just being varying serializations thereof. It also allows the website to show what has been amended — no need for dumping stuff to GitHub and then diffing it! (e.g., see the annotation on 28(1)c in http://www.legislation.gov.uk/ukpga/1998/29/part/IV)

That all said, there tends to be a delay between the PDF being uploaded and everything being marked up and entered into the RDF database (see the "new legislation" on the home page).

I actually wrote a tumblr post about that a while ago. It also included a user engagement aspect that would allow users to subscribe to topics of interest to receive a curated feed of changes that could each have a dialogue/poll emerge. I hadn't really thought through the specifics but its awesome to know that something like this is already started to some degree. Thanks for that!

Here's the post if interested.

http://proggr.tumblr.com/post/52801512630/githubberment

In France, all is on the Web, but you’re not allowed to crawl it unless you pay :/

This is currently done via scraping:

https://github.com/divegeek/uscode

The diffs are huge.

It would be even awesome-er (and more useful) if you could parse individual bills and amendments into diffs, which get merged into 'master' as they become law.

I'd love to `git blame` the U.S. code.

I was under the impression law is essentially "append only". New laws override existing laws, but the text of the existing law never changes.

We actually have something similar in germany, called the "Bundesgit" (https://github.com/bundestag/gesetze)

"Diffing the law" between different date points would be amazing. I hope you follow your temptation in this regard.

I'd do it myself but I'm already neck deep in work and volunteer projects outside of work.

The funny thing is that a lot of law is structured like definitions, the actual law, then consequences. Any of those three can change independently, and change the meaning of the law. So diffs are often not as useful as you wish.

I just created a repo with all the codes[1]. There was a convenient link to all codes in one zip file. I will note that this is a massive amount of text.

There is a link to the schema used and a stylesheet (I assume for the xhmtl maybe?) that I would like to add in. But one step at a time.

[1] https://github.com/varikin/UnitedStatesCode/

Please do! This is the kind of thing that hatches in my mind as a great idea, but withers due to a total lack of follow-up.

Probably more important than just diffs is actually a dependency graph and a topical index. While they've tried to do this via titles/chapters and references, linear breaks are never going to be as successful as many-to-many linkages as every Bible concordance out there demonstrates.

I'd really like to see a graph of the number of bytes over time, and what sort of curve it fits.

NZ laws on github: https://github.com/Br3nda/legislation/tree/master/act/public

"In short, that allows a lawyer to answer almost any question with "it depends," and start billing. "

No, this is usually "lack of sufficient facts to answer question".

If a crime has 4 elements, and you've said you've committed 3 and maybe the 4th depending on how you look at it, the answer to the question "can i be convicted" is "it depends", because it depends on the argument about the 4th element.

If your followup is to claim that there should never be questions about whether you've achieved some element of a crime, it should be binary, that's simply impossible.

Crimes, for example, are often about intent and causation. Until science provides us with historical mind reading devices, and completely accurate point-in-time reconstruction of the world, we're going to have to rely on various theories and evidence (much like science relies on theories and evidence).

Even point-in-time reconstruction wouldn't solve all causation issues, because the issue is often a fairness one: Is it fair to hold the defendant liable for a crime X steps removed from his own action.

With respect, only someone who hasn't studied law could say this.

Law is complicated and filled with ambiguity because the world is complicated and filled with ambiguity.

Unlike lawyers, we can simply will some of that away. We can make an enum which smooshes together disparate phenomena, that feeds a dropdown list and require the customers to use it. We can ignore very rare corner cases and include a comments field. We can do a lot of things to simply make essential complexity go away while we bicker amongst ourselves about accidental complexity.

Law can't do that. Law must treat all cases, all combinations, as they arise. Every single day a configuration of people, motives and circumstances emerges that has never before emerged, could not have been foreseen and which may never occur again. The law must still apply.

Law is hard because the problem domain is everything humans do.

Let me know how many story points you think that scope is worth.

Interpretation isn't an unguided, irrational process. When your lawyer answers "it depends" its because you've placed him at one point in what's really a decision tree and asked him: "where does it end?" It's like asking a programmer whether a particular token in a C++ program is a type name or an identifier. It depends.

If you imagine the universe of human activity as a multi-dimensional space, you can think of laws as carving out subspaces with lines (that may be more or less fuzzy) and attaching significance to particular subspaces. If I ask: "my x coordinate is 5, where am I?" what do you answer?

You make this sound like a bad thing. Would you really prefer "black and white" laws that leave zero room for flexibility or interpretation in light of a given situation?

It doesn't matter as much as you say. The fact is, at the very least we can now know better where the fenceposts are, and this will have the undeniable effect of creating a more educated citizenry. A secondary effect is that motivated people can now create tools that reveal ways in which legislators are positioned in relation to those fence posts, and why their interpretations (such as they may be rationalized in accordance with your observation regarding law enforcement) are positioned where they are.

Perhaps most importantly, this will serve to make the law shallower. Currently, many laws are passed with large parts in reference to other laws, definitions from 400 pages elsewhere in the text (themselves with referential connections to disparate locations), and providing a way to analyze the law programmatically will have inevitable analytical benefits that allow citizens to know exactly how the sausage is made, whose farm the animals are coming from, and how it's been spiced.

There is this wonderful article by John Hasnas, "The Myth of the rule of law". It brings a number of very interesting ideas:

* The law is inconsistent and possess contradictory rules. From contradictory premises, sound arguments

* It gives concrete examples of legal cases and how completely opposite judgments could be rendered, both being based on correct use of laws and analysis.

* Laws are not applied impartially, are not neutral, apolitical nor objectively applicable by judges. People who believe they live under "a government of laws and not people" tend to view their nation's legal system as objective and impartial. Obey and subject themselves.

* There is a cognitive dissonance between the laws and the people who make them. People generally believe that the law are generally good and should be obeyed. People generally believe that politicians tend to be corrupt and influenced. The people who make the law cannot be trusted but yet the laws they write should be.

At least for me, accustomed to logic, determinism and mathematical models for laws of nature, this was actually a surprise. Instinctively I still find it distasteful, though I can't say I have anything better to offer other than the unrealistic ideal that all judges would be guaranteed to be truly impartial. I have never directly benefited nor been harmed by the nature of this 'code' either, unless you count getting out of traffic tickets by switching on a learned persona that has people skills.

No, judges (who mainly interpret words in laws and partially juries), are not allowed to just make things up all the time, which is what you're implying.

I can't believe that comments like this get upvoted by the same people who decry the practice of viewing programmers as monkeys or typists.

Good luck starting your own country with a legal system that doesn't require interpretation, or heaven forbid, lawyers.

HN's InclinedPlane put this sad reality of the legal system into perspective wonderfully: http://news.ycombinator.com/item?id=4753117

If I used a LinkedList in my code instead of an Array will it speed things up?

It depends...

The US legal system is case law. of course, that makes interpretation even more non-deterministic, as you depend on your $10,000/hr legal team to dig up favorable ruleings

What you've described is a trait of human beings, not politicians and lawyers.

Actually, you wouldn't. This is the US code, not the legislative info :)

I actually tried to create this once (with a team behind me, in fact) with what's available or possible with the legislative info.

THOMAS theoretically published in XML, but it's missing a lot of info.

Not only is this info not published, it's not even stored. They still are literally passing bills around to each other in some cases. You'd have to sit in on committee markup hearings, etc.

Even the small amount that is published from markups or whatever doesn't tell you who did it, only that it was done.

Some committees were willing to give more info. None were willing to essentially publish the in-flight staff attorney or other versions that would tell you for real who changed something.

Remember also that hg/git diff does not display what really happened.

It is displaying "how can i produce version B from version A using some minimal or near-minimal set of changes", converted to a line (or sub-line) based set of changes.

This does not tell you history, only one possible set of changes.

THOMAS does publish some in-flight versions of bills, but again, it's not really enough to do what you really want to do. I can tell you a bill changed between introduction and markup, and was different again when it got back to the floor. I can even, in some cases, tell you what was amended/deleted. I can't tell you who did it.

(Well, i can tell you with some percentage accuracy, because we built a machine learning model, but ..)

You can already do this - Congressional voting records are publicly available. However, even if you correlated these records with specific legislation, you would be making a dangerous mistake by missing the big picture.

Many laws are passed jn bundles and require compromise. Just because your Senator voted to support farm subsidies, doesn't mean he likes them - it just means he wanted a minimum wage increase to pass more.

Hopefully, as a result, the disturbing statements like the ones below will become obsolete:

James Duane, Regent Law School professor, defense attorney:

"Estimates of the current size of the body of federal criminal law vary. It has been reported that the Congressional Research Service cannot even count the current number of federal crimes. These laws are scattered in over 50 titles of the United States Code, encompassing roughly 27,000 pages. Worse yet, the statutory code sections often incorporate, by reference, the provisions and sanctions of administrative regulations promulgated by various regulatory agencies under congressional authorization. Estimates of how many such regulations exist are even less well settled, but the ABA thinks there are ”nearly 10,000.” http://youtu.be/6wXkI4t7nuc?t=5m18s

Supreme Court Justice Breyer:

"the complexity of modern federal criminal law, codified in several thousand sections of the United States Code and the virtually infinite variety of factual circumstances that might trigger an investigation into a possible violation of the law, make it difficult for anyone to know, in advance, just when a particular set of statements might later appear (to a prosecutor) to be relevant to some such investigation." http://www.law.cornell.edu/supct/html/98-93.ZD.html

You can look at relatively small jurisdictions and copy what's appropriate. Throw out hunting and fishing and take Alaska's laws in 1959 and you've got the minimum for a jurisdiction. Same for the Canal Zone. You can add legislation from other jurisdictions that works, like New Zealand's torrens (not torrent) system. Alberta's condo laws. I wrote a set of laws for a new jurisdiction, there's no reason to start from scratch.

Usually countries inherit the laws from the country they are decended from.

Just from a Natural Language perspective, this is amazing. By using the code as as a corpus, we can see how differently legal language differs from day-to-day parlance. Such an analysis might be fascinating, especially if we compare to word frequency in Google Ngram's historical index. I suspect legal language trails several decades behind modern lingo.

Semi-intelligent queries can be executed, such that ignorance of the law might be abated. Imagine saying "Siri, is it illegal to do X?" and Siri answering you. This is important, because of the "Three Felonies a Day" syndrome with unwitting violations of the law. http://www.amazon.com/Three-Felonies-Day-Target-Innocent/dp/...

The law can be interpreted as a set of rules matched against an action in a particular context to determine whether said action is legal.

Right now lots of people do this manually.

Maybe computers can do it better.

If computers can do it better, maybe people could focus on writing new laws and refactoring old ones instead of repeatedly interpreting old laws for each case.

Similar question and answer on another site: http://www.reddit.com/r/technology/comments/1jfufh/the_entir...

In practice, it probably doesn't make any difference. Assuming, for the sake of argument, that XSD is expressive enough to allow the authors to say what they're trying to say, in an unambiguous way, it's more or less a moot point. RELAX/NG is awesome, sure... but to some extent the "war" between it and XSD is a religious war.

Pretty much every popular & widely used language / platform has XSD support... hell, it's so commonplace there are probably schema aware xml parsers in Brainfuck, INTERCAL and Befunge.

RELAX NG may well be a better choice, but saying this is "tragic" strikes me as a bit of excessive hyperbole.

Maybe try to organize some tech-minded staffers on the hill? Get them to put bills onto a github repo when they are made public, maybe even XMLizing it to make it more exposed to machine analysis. It'd be a huge step in congressional transparency, especially now that bills are going into the high hundreds and thousands of pages (not a dig at the healthcare bill in particular, just a trend).

Imagine if it was done essentially as a "pull request" in the future. That would be amazing.

The encoding looks like valid UTF-8, at least for the first few pages that I glanced at.

I did notice the section references look a little strange in vim, e.g. "act Aug. 10, 1956, ch. 1041, § 1" near the top; it consists of c2 a7 (section sign), which looks fine, followed by e2 80 af (narrow no-break space), which shows up as a box in vim.

Yes, old name is: Title 38 - Pensions, Bonuses, and Veterans' Relief

http://www.gpo.gov/fdsys/pkg/CFR-2004-title38-vol1/content-d...

I'm more annoyed that they are out of order with reference to which clause in Article I, section 8 their existence is justified by - but I suspect there's a reason for that.

Congress has a site for it.

http://uscode.house.gov/search/criteria.shtml

This is only the codified portion of the federal law. So called "Statues at Large" that are passed by Congress but not codified into USC are not included. Plus this does not include the federal common law based on the rulings of federal courts.

As the schema guide explains, in 1999 they picked XML based on the study from 1996:

"Following a 1999 feasibility study on XML/SGML, the Committee on House Administration adopted XML as a data standard for the exchange of legislative documents" http://uscodebeta.house.gov/download/resources/USLM-User-Gui...

It took the US Govt 17 years to release 200,000 pages of the US code in XML.

Considering typical timescales for this sort of projects, i'd say we should be happy it's not in CSV or SGML.

Why? JSON should just be for data transfers. XML is for document formats. While the lines get seriously blurred that is there general use.

I hope you're joking.

XML is a superior format for semantically marking up content from a very detail-oriented dataset like this.

Given the level of detail in the XML, I shudder to think what monstrosity that would create.

Heck they didn't even bother to include a way to inherit styles, they are just included in every case.

Time to write a json wrapper. I nominate you!

i'd prefer markdown /sarcasm