HNHacker News
TopNewBestAskShowJobs

drewcrawford

4,834 karma · joined June 13, 2008

AI tooling and systems/SDK engineer · Rust, WebAssembly, Swift/Apple platforms · wasm-bindgen contributor · building AI Bot Mafia (botmafia.games) · sealedabstract.com · github.com/drewcrawford
submissionscomments
drewcrawford··on Don’t Mess with The Google
Yeah, very legal.

The "right of first sale" means that you can't be sued (well, you can be sued but they would lose) if you sell a book you own. It doesn't say that the publisher of that book must continue to sell you additional copies or that the author won't give you a dirty look if you pass in the street or that someone may take steps that don't involve suing you for copyright infringement to discourage your activity.

This is, by the way, the same trick that powers RHEL – the binaries are all GPL, so you can totally redistribute them to anyone. If you do though, RHEL will cancel your account.

You own your data but not the servers on which that data is stored, so you can't compel Google to give you access to it, any more than you can compel me to produce your comment from my browser cache if you lose it. Refusing to give you access to their servers definitely isn't theft.

Very legal. Arguably unethical, but that's a different question.

drewcrawford··on White House urges ban on non-compete agreements for many workers
> 1) deontological: freedom to contract is always a good thing.

A good thing for whom? It is good for lawyers and others who benefit from writing contracts.

Imagine for a moment a world in which you sign a non-compete before ordering a burger. It sounds silly, but sillier clauses are signed in order to purchase far cheaper consumer goods every day.

Is the world really a better place because two parties are hypothetically able to achieve their longtime dream of finally combining a legal instrument and a sandwich? Are you excited that you "have local knowledge" and are ready to exploit this knowledge to review legal terms with counsel before ordering your burger OR is this hypothetical benefit dwarfed by a very real huge volume of people who aren't going to hire a lawyer to negotiate a contract of adhesion for a $10 purchase when under social pressure while they're dining with someone else?

Freedom to contract is not always a social good. It introduces complexity, and complexity creates friction which prevents people from participating optimally in a market and prevents the market from clearing. Contract innovation is not optimal for the burger market.

What is optimal is common law, which is mostly the same everywhere so that an entire society understands how it works. It's how the police know what to do when facing a diner who didn't pay their bill, it's how you know what to do when the order is wrong, it's how a business can hire employees who know all this stuff without training them on custom procedure. Standardization is optimal for the vast, vast majority of transactions, so that people can get on with whatever they do when they're not negotiating legal instruments. The benefit of designing something bespoke for some marginal case is dwarfed by the cost of increased complexity of the transaction and the deviation from the established standard.

If you believe that job contracts are so positive and innovative, insist on reviewing one as a condition of doing an interview. If it is really about local knowledge, surely it will be positive for all involved to develop that knowledge early and comparison shop. In reality people will look at you funny and probably decline, because the real function of that contract is to appear after you've decided to take the job and exploit social pressure in order to get a stronger deal for the company.

drewcrawford··on New 'Dead Sea Scrolls' Revealed
Well, OP is confused. There are of course some differences between the MT and the Dead Sea Scrolls in Daniel, but they are very minor [0], to the extent that in many cases the MT actually harmonizes better with the scrolls than they do with each other.

What OP may be referencing are the much more significant variations between the MT and the LXX. The LXX version of Daniel is quite a lot longer, uses very different details, and even includes additional stories [1] not present in the MT.

This has presented something of a problem, since although the LXX is a translation of a source text and so is inherently inferior to it (like any English translation would be), it is a translation so incredibly old that the later parts of the bible use it when they quote the earlier parts, and it was the standard version in the ancient world for quite some time. Meanwhile, the MT is an actual manuscript in the source language but arrives at least a thousand years later on. This creates a bit of a puzzler about which one to follow when they are quite different, and they are quite different in Daniel.

The DSS basically settled the issue with regards to Daniel in favor of the MT and against the LXX. There was not really anything "new" that we did not have in the MT, but there were some LXXisms that fell out of favor.

However, in other books it went the other way, sometimes favoring the LXX reading. This suggests that the LXX and the MT may reflect different underlying source texts, that are both quite old and both have some legitimate claim to being authoritative. The situation is complicated somewhat because most of these books are more compilations than they are a work by a single author, and deciding where the original work ends and the interpolations begin is pretty arbitrary.

[0] https://books.google.com/books?id=6f9-Fnmy3sEC&pg=PA149&lpg=...

[1] https://en.wikipedia.org/wiki/Bel_and_the_Dragon

drewcrawford··on New 'Dead Sea Scrolls' Revealed
50,000 people is the order of magnitude of the Israelites alive at this time. So if 50,070 people died, Israel would cease to exist as a geopolitical entity, whereas this text has them going on to win a war against Philistia merely 10 verses later.

But leaving that aside, a reading of 50,070 has a number of other problems. First, it does not explain how such an exact figure was calculated, which would have been a major statistical undertaking at this time. Compare to a similar situation in Numbers 1, where nearly 60 verses are devoted to explaining the procedure used to arrive at figures much less exact than these. (Those figures are difficult too, but not because of their exactness.)

But leaving that aside also, it doesn't account for the glaring error in the number itself: שִׁבְעִ֣ים אִ֔ישׁ חֲמִשִּׁ֥ים אֶ֖לֶף אִ֑ישׁ is a bizarre way to write a number. It's like saying my software has "70 lines 50KLOC". Everything is wrong: the place order (fifty-thousand seventy, not seventy fifty-thousand), the conjunction (fifty-thousand and seventy), the noun ("lines") is redundant and even interrupts the number. If somebody said "70 lines 50KLOC" you would not assume they meant 50,070, you would assume they made an error.

All lines of evidence point to some problem with the figure, but there is really no agreement as to how it is resolved. The modern style is to assume that "70 men" and "50,000 men" are two different figures, one of them is spurious, and choosing 70 is better than choosing 50,000. The style falling out of favor is to assume two figures for two distinct things, one of which is lost, which is how you wind up with the "70 men and 50 oxen" readings.

Finding a new manuscript could very well blow the lid off this debate, but it would have to be quite old, as both the problem and the debate around it are positively ancient. I believe we have the 50,070 reading all the way back to 50BCE, and we certainly have rabbinical sources arguing about it by c. 300. The book itself only dates to 550 BCE, so it is a very narrow window to stumble onto a manuscript.

There is an outside chance this particular section may be lifted from an older, lost work all the way back to the time of David, so this problem may even pre-date the book in which it appears. Unfortunately we have only tens of lines of manuscripts from that whole period, so recovering that work – if there even is an earlier work, which is not at all clear – is terribly unlikely.

drewcrawford··on New 'Dead Sea Scrolls' Revealed
Most of the textual differences are ironed out in the source language which are not especially accessible. English bibles are not translated directly from manuscripts but from one or more "critical texts" which are scholarly reconstructions of the underlying manuscripts. A common one is Nestle-Aland [0], you can read it online. The smaller text appearing there are variant readings that have some degree of manuscript support.

The vast majority of them are terribly boring, by way of example, the very first one on the page:

> τοῦτο δὲ ὅλον γέγονεν ἵνα πληρωθῇ τὸ ῥηθὲν ὑπὸ κυρίου διὰ τοῦ προφήτου λέγοντος ἰδοὺ ἡ παρθένος ἐν γαστρὶ ἕξει καὶ τέξεται υἱόν, [καλέσουσιν τὸ ὄνομα αὐτοῦ Ἐμμανουήλ] ὅ ἐστιν μεθερμηνευόμενον [μεθ’ ἡμῶν ὁ θεός.]

> All this took place to fulfill what the Lord had spoken by the prophet: Behold, the virgin in the womb will conceive, and will bear a son, and [they will call the name of him Immanuel], which is translated, '[with us God]'

The [bracketed] parts are omitted in some manuscripts. These omissions may feel significant, but since this is a quotation of well-known saying, the shorter readings are probably some scribe trying to scrimp on paper who thinks you should go buy that book if you wanted to know what's in it. This can be annoying in cases where the book in question didn't make it out to modern scholarship, and all the scribes scrimped on paper that day, which does happen.

Less commonly (although not infrequently), a textual problem is significant enough to bleed all the way through to the end user. This typically happens when there is no agreement about how to resolve the conflicts or the possibilities are so strange as to be nonsensical. One amusing example is 1 sam 6:19 [1]. There is overwhelming manuscript support for a reading like this:

> But the Lord struck down some of the people of Beth Shemesh because they had looked into the ark of the Lord; he struck down 50,070 of the men. [NET]

However, in addition to there being a glaring grammar error in the manuscript, this reading presents a number of obvious problems: logistical (how exactly do you get 50,070 people to all look into a box?), numerical (this is a surprisingly exact figure to collect during wartime in enemy territory), internal (elsewhere in the same book Beth Shemesh is a tiny rural village), archeological (it really was a tiny rural village).

This problem has lead to various wildly different resolutions over the years, a small sampler:

> But God struck down some of the inhabitants of Beth Shemesh, putting seventy of them to death because they looked into the ark of the LORD. The people mourned because of the heavy blow the LORD had dealt them. [NIV]

> God struck down the men of Beth-shemesh because they looked inside the ark of the LORD. He struck down 70 men out of 50,000 men. The people mourned because the LORD struck them with a great slaughter. [HCSB]

> Then God smote those of Bethshemesh because they had looked at the ark of the LORD; he smote fifty thousand of the people and seventy principal men. And the people lamented because the LORD had smitten the people with such a great slaughter. [Jubilee]

> And He smiteth among the men of Beth-Shemesh, for they looked into the ark of Jehovah, yea, He smiteth among the people seventy men -- fifty chief men; and the people mourn, because Jehovah smote among the people -- a great smiting. [YLT]

> Perhaps the text should be understood to read the LORD killed 70 men and 50 oxen. [NKJ footnote]

[0] http://www.nestle-aland.com/en/read-na28-online/

[1] http://biblehub.com/1_samuel/6-19.htm

drewcrawford··on Opposition to Galileo was scientific, not just religious
You are equating knowledge with mere numerical literacy, e.g., you are imagining that knowledge is a set of fundamental constants to be measured, and that he who knows the constants to the highest precision is the greatest.

In fact knowledge is the thing that allows you to choose which constants to calculate at all. It is not intuitive why a light in the sky should have a size. What is the size of a sunset? What is the size of a rainbow? What is the size of an eclipse?

Knowledge is not the mere accumulation of measurement. It is the organizational structure that allows you to separate between what we can and cannot measure, what we have measured before and have not measured yet. It is what tells us whether "the" sun is a singleton or that "a" rainbow today is not the same today as tomorrow.

> who don't know history

I think you would benefit from reading up on the current consensus among historians on the Conflict Thesis before presenting it to those of us who have caught up on our Ferngren, Barker, etc.

drewcrawford··on Theo de Raadt on LLVM license change
> The Apple EULA objectively has clauses in it that state that users must follow the "acceptable usage policy" and it is enforced regardless of any external agreements -- this would mean that Apple would be violating the GPL if they distribute any GPL software in their store.

And Red Hat EULA objectively has clauses in it that you may not redistribute RHEL.

There is a slippery slope here, the FSF has decided for whatever reason that RHEL's restrictions "weren’t really restrictive, because they were so easy to comply with" [0] but there is really nothing "objective" about that determination.

[0] https://www.gnu.org/licenses/quick-guide-gplv3.html

drewcrawford··on A disaster is looming for American men
> If your theory stands true, then how come that overall, the global economy is growing? Doesn't it mean that we've created more jobs that we've destroyed?

The number of jobs does not correlate to GDP growth [0]. So whether "the economy is growing" is not terribly related to whether jobs are being created. This is somewhat puzzling under Adam-Smith-Wealth-of-Nations ideas about economics, but it's an empirical fact.

> Technological disruption is not new. It's been around forever.

You seem to be misunderstanding the argument; I'm not arguing for something new. I'm arguing that negative effects of technological shock are well-attested in the historical record: you can look back at historical data and plainly see them.

[0] http://www.nytimes.com/2016/01/18/business/economy/us-growth...

drewcrawford··on Theo de Raadt on LLVM license change
There is debate about to what extent patent protection is already extant in the current licenses. One example, the MIT has

> to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software

The word "patent" does not appear here, but the words "use" and "distribute" seem to imply the granting of rights necessary to achieve those ends (of which a limited patent license would be one), and the phrase "deal in the software without restriction" is a statement so broad that it seems difficult to argue a patent license (and perhaps many future IP innovations unknown to us today) was not granted by the rules of equitable estoppel.

The reason for this license change is to resolve the ambiguity that arises because the word "patent" does not appear as a word in this clause. It was found that the GPLv2 did not grant patent rights because it was not mentioned in that license [0], although the circumstances of that case and the text of the MIT are far apart.

As a result of all this LLVM may go around and get people to sign off on the Apache grant but it is likely to be under the banner of "clarifying" an implied patent grant rather than establishing a new one. Depending on how worried the lawyers are about the legal meaning of "deal in the Software without restriction" they may not be worried about the contributors they are unable to track down.

[0] See https://www.fenwick.com/FenwickDocuments/potential_defenses.... for an analysis of the GPLv2 situation

drewcrawford··on Theo de Raadt on LLVM license change
Well to cite an actual IP attorney [0], yes. A sublicense means that you (the licensee / sublicensor) grant somebody else (the sublicensee) the right to use the software, and you can do it under any terms you are permitted by the grant of sublicense.

The MIT/NCSA do place some restrictions on sublicenses ("subject to the following conditions") but I don't see any reason why the Apache license would present a problem as the choice of sublicense.

I am not a lawyer, I am not your lawyer, this is not legal advice.

[0] http://danashultz.com/blog/2010/10/19/what-is-a-sublicense-a...

drewcrawford··on Theo de Raadt on LLVM license change
"Additional restrictions" can be funny. You may recall the FSF's position that the app store EULA placed "additional restrictions" on the GPL, a position that was and remains controversial.

In this case, it is a little harder to understand OpenBSD's interpretation of additional restrictions because they haven't listed their objections in any centralized place, but from reading around, they seem to believe:

* "You must cause any modified files to carry prominent notices stating that You changed the files" is an additional restriction * "any Contribution intentionally submitted for inclusion in the Work by You to the Licensor shall be under the terms and conditions of this License" is an additional restriction * The sheer length and complexity of the Apache license suggests to them there are other additional restrictions buried in there somewhere

drewcrawford··on Theo de Raadt on LLVM license change
Both of the licenses used in LLVM contain the following clause:

> Permission is hereby granted... to deal with the Software without restriction, including... the rights to... sublicense

The effect of this clause is that all LLVM contributors have already consented to a chance in license, whether that change is done by you, me, Debian, or the Pope.

The only question is whether the LLVM organization wishes to exercise this authority itself. It's clear that the majority of the key contributors are in favor of it, so it seems like a foregone conclusion at this point.

drewcrawford··on A disaster is looming for American men
It's not actually any different; "technical unemployment" has caused many problems that are clearly visible from a cursory examination of the historical record.

The obvious place to start is the horse: the horse population in the US peaked at 26M in 1917 and had declined to 3M by 1960, almost 1/10th of its peak value [0].

When those horses were unemployed by the automobile, we did not retrain them in new lines of work. They just died. It got so bad that conversation societies were established to protect endangered breeds.

Eventually we created a handful of new jobs in the recreational sector for them, so the population of horses is rising now, but it is nowhere close to the peak value, and of course we did not create those jobs within the lifetime of any displaced horse.

In fact what the historical record shows is that massive technological shocks produce long-term unemployment. It doesn't seem to matter if it's a horse, because of course they're just resources to be exploited in the furtherance of human goals. But then again, so are you.

[0] http://www.humanesociety.org/assets/pdfs/hsp/soaiv_07_ch10.p...

drewcrawford··on Continuous Intrusion – Why CI tools are an attacker's best friends (2015) [pdf]
Heavy Gitlab CI user here. The real problem is isolating the runners themselves. You can assign runners to individual projects for isolation, but when the runners can't be AWSed (iOS hardware, for example) you have to make tough choices about buying more stuff to sit under your desk right next to the stuff you could be using that's idle.

IMO some basic isolation measures like clearing the cache between projects would be good. Killing active processes when a job ends.

drewcrawford··on Canadian Comedian Fined $42,000 for Telling a Joke
I did read it. I do think it is a travesty.

It seems unfortunate to me that we can't have a civil debate about important issues without importing baggage like "right-wing misogyny". Perhaps that is a good reason to flag the story, but if so it seems like more of a bug than a feature.

drewcrawford··on “Please note, the hashtag is for our paying advertisers”
Then it may surprise you to learn I still live in the South :-)

I understand how it could be read as a bit of a partisan attack, particularly around this time of year. In reality though, I think every place has its own dysfunction. It's just that Edmond's particular dysfunction is both important to understanding the story, and very foreign to HN readers, whereas detailing the dysfunctions of other areas might "provide balance" but would not help anyone make sense of this story.

But to provide balance anyway, Edmond is very different than most people's idea of "red area". The population is very highly educated. The public school system is universally loved, and there's even a public university that is excellent. So when you hear someone express their hatred for "big government" it is in the context of empowering a local government that provides actual services that are very competently administrated, not, e.g. a desire to burn everything down.

It is also one of the only semi-urban areas I can think of where minimum wage is definitely a living wage, jobs are available to anyone who wants to work, "corporation" means your next-door neighbor's 10-employee plumbing company, etc. So I think an argument can be made that "red" policies make much more sense there than otherwise expected.

It is really the sort of place that produces a William Buckley [0]. It's a pragmatic, intellectual, moderate branch of conservatism, that is growing concerned about its own extinction. Based on this election, I think it's a pretty justified fear.

[0] https://en.wikipedia.org/wiki/William_F._Buckley_Jr.

drewcrawford··on “Please note, the hashtag is for our paying advertisers”
I grew up in this city, so it is surreal to see it on HN.

Edmond is a special place that you must understand to make any sense of this story. It is effectively Republican Heaven. It is a small town, well-eduated, white, wealthy, growing, and suburban. Lots of small business owners and people who have made their own $1M, which they naturally achieved through some combination of hard work and attending church regularly so that your customers and business partners will trust you. Cost of living is shockingly low. All the economic, racial, and social issues in our country are happening somewhere else very far away, and the biggest threat Edmond faces is that the faraway people may and/or already have begun to change things and therefore ruin Republican Heaven.

This company exists in that context, capitalizing on a high concentration of small business owners, offering them what effectively amounts to advertising services that theoretically convert into walkins to your hair salon or whatever.

Edmond has a strange relationship with social media. On the one hand, social media is big business, and big business is always, always good. On the other hand, social media was created by a bunch of faraway Blue-State Liberal Commies. So in general they would be one part in favor of exploiting it, one part completely ignorant about how to do that, and one part determined to do it differently than the Liberal Commies do.

All of that to say, forming a cartel around a hashtag is an entirely plausible business model for Edmond. If a business used it without paying, well, that's cheating, and cheating is a sin, and you'll never sin your way to making your $1M. All you have to do is put out a passive aggressive note reminding people you have a trademark or something and the bad actors will be sufficiently shamed to change their ways.

And the social media backlash is, obviously, just a bunch of Blue-State Liberal Commies who live too far to visit your hair salon anyway, so nobody cares about them or their opinions.

The real problem this company faces is proving that a hashtag translates into increased hair salon appointments, which is the kind of problem Edmond actually cares about. Small business owners take much more seriously than SV culture the concept of value for money. If it's not effective advertising it will be dead in the water, with or without any help from outside.

drewcrawford··on On the boundaries of GPL enforcement
Well yes, and as a developer I am very sympathetic to the "author's right" argument.

But we must also allow that the netfilter people have the right to decide who is and who isn't one of their core developers, and publish their philosophical positions on their own mailing lists. So I am not sure how the "author's right" angle would assist us other than to say that everybody can, and in this case did, exercise their rights.

Perhaps a sounder analysis could begin on the basis that ideally, people would work together, and in order to work together, they must at least be able to talk to each other. It seems unfortunate to me that this person didn't engage in the dialogue that these people apparently sought, even to say that he objected to their position.

Then again, it is possible that this developer saw no upside in revealing his position before a potentially hostile audience, and perhaps removing him from the project in some way served as validation of that concern.

I do think one of the things we are losing as a society is the ability to tolerate different points of view. I think that might be the real story here.

In a simpler time there was merely a battle between free and proprietary software, but we seem to have decided that is not a sufficient set of tribes in which to divide ourselves. We must have copyleft vs permissive, and the copylefters must split into the watered-down LGPLists vs the full strength GPLists (vs the espresso-grade AGPLists). We must have a battle between the One True GPL2 vs the Revised Third Edition. And now apparently we need to sign Enforcement Principles to fix all the things we forgot in the licenses themselves.

I guess I understand the motivation to continue to do this, but when our differences are so large as to prevent people from even engaging each other constructively I find that unfortunate.

drewcrawford··on A Report on the Flawed 2016 Democratic Primaries
It is very clear to anyone who wants to look into it that Russia is the culprit. There were two attackers; one used the same tradecraft, malware, and command-and-control servers as the Sofacy group, [0] which we have known to be Russian intelligence over dozens of large attacks all the way back to 2007. The other attacker used the same tradecraft, malware, and command-and-control servers as the "Cozy" group [1] which we know less about but reuses the same technology as other Russian intelligence actors and attacks Russian targets. As far as perpetrators, we definitely have the right people.

Motive is less clear. In reality, they probably hacked the DNC because everybody hacks everybody; that's the reality of modern cyberespionage. America hacks the Russian political parties, they hack us back. But this time someone screwed up and they got caught, so now we need a cover story.

The coverstory of trying to manipulate the election is basically convenient for all involved. It is far fetched enough to sound unbelievable, so it provides diplomatic cover. It also provides cover for why there's no similar story at the RNC, which is better than admitting either the Russians couldn't do it, or they could and they got away with that one. It provides cover to Clinton who can now claim the Russians want to elect Trump so please focus on that instead of anything to do with email, and it provides cover to Trump who would like to keep the word email in the news. This is the story that all our characters are happy to go with.

But the Russians being the perpetrators? That part is definitely true. There is way too much evidence over way too many years. The DNC did not create a Russian intelligence organization in 2007 in order to cover up their email leak in this election.

[0] https://en.wikipedia.org/wiki/Sofacy_Group

[1] http://securityaffairs.co/wordpress/36195/cyber-crime/cozydu...

drewcrawford··on GitLab 8.10 Released with Wildcard Branch Protection and Manual Actions for CI
I have it set to 2 already :-) Unfortunately I still ENOMEM at least once a day.
drewcrawford··on GitLab 8.10 Released with Wildcard Branch Protection and Manual Actions for CI
Hi sytse, while I've got you.

I've been annoyed these last few releases by high memory usage. Every day, my backups fail due to ENOMEM.

I have adjusted unicorn-worker-killer to undo the increased memory usage you did a few releases ago:

    unicorn['worker_memory_limit_min'] = "300*(1024**2)"
    unicorn['worker_memory_limit_max'] = "330*(1024**2)"
(8.9.0's default is a totally absurd 400-650MB.)

and while that stops my server from falling over in the first 10 minutes, it does consistently fall over once a day.

Here is a screenshot of my htop, I'm not sure where to look next. 25% free sounds okay, but it's not enough to run a backup, or even run `gitlab-rake`.

http://weblinksdca.s3.amazonaws.com/Screen%20Shot%202016-07-...

drewcrawford··on Lepton image compression: saving 22% losslessly from images at 15MB/s
I think you may be crunching the numbers the wrong way...

Dropbox is valued at $10b (regardless of what you think of that number, someone was willing to buy at that price). But they only have 1800 employees. If by some miracle 80% of them are engineers that is $7M per engineer. All this for a "pure software" business. No inventory, no manufacturing, they only recently started doing their own operations.

Investors would be telling them "more engineers" => "more software" => "better Dropbox". So now you have to go out and snap up engineers in Silicon Valley, which is very very hard. But you don't offer the perk of a Google/Apple line on a resume nor the compensation of e.g. Microsoft. What do you do?

Answer: you offer people the chance to work on cutting edge breakthrough technology. Not only that but it's all open source! This is a dream job for some people. They will turn down every other gig for this one.

It doesn't matter that PB/mo is chump change because an investor doesn't know that, they only know the engineering headcount. If by some miracle an investor does know this is a waste of time, then you just point to this HN post and observe how many developers are interested in this technology and how it is attracting developer mindshare that can be exploited for additional hires down the road.

I'm not saying it's rational–it's not. But I think there were strong incentives to greenlight a project like this, even if the actual cost savings were zero.

drewcrawford··on A Saudi Arabia Telecom's Surveillance Pitch (2013)
spiped is GREAT tech, and shares a lot of the same philosophy: small implementation, simplified protocol, no negotiation, key pinning, etc. It is really, really solid for the problem it solves.

However, it's designed around a simplified threat model. In spiped, you use pre-shared keys (the same key on client and server), that "somehow" you got a copy onto both systems without anyone seeing them. But you can't just drop your key into a mobile app binary, submit to the app store, and expect people not to reverse-engineer it and start reading other people's traffic. So now you need to figure out how to derive spiped keys on a per-session basis in a cryptographer-approved way™ and distribute them securely and verify the key someone told you to use is not actually from the NSA, and then on top of that think about perfect forward secrecy and certificate pinning and various other "beyond TLS" guarantees.

There are solutions to all those problems, tarsnap ships really cool solutions, but we are not talking about spiped anymore, we're talking about spiped and a long list of other things, and the other things are easy to get wrong if your name is not cpervia.

Another practical problem with spiped is the performance degrades for small messages. This doesn't matter if your use case is doing backups, but if your use case is "I typed the letter 'c' in the search bar and where are my search results? I thought this was supposed to be a performance technology" it matters a LOT.

drewcrawford··on A Saudi Arabia Telecom's Surveillance Pitch (2013)
Noise is not really a protocol but a toolkit to build one. It's really a symptom of the problem. We expect individual developers to take these tools and figure it out, when they won't because they don't care enough. If by some miracle they care they will probably screw it up unless they like reading crypto papers. It also has various practical problems (e.g. the 64k message limit) that you have to deal with some way, probably at the expense of performance which is a useful vehicle to sell security technology, possibly at the expense of security (e.g. early session termination).

QUIC is more interesting and I'm evaluating if we can deploy a modified form of it into our transport layer. The key problem with QUIC alone is that in practice many networks block UDP traffic, so you need a TCP fallback solution so customers don't complain that every other app works except yours. But when your threat model is "government compels certificates" then you can't fallback to TLS. The government will just block UDP traffic at the router, downgrade you to TLS, and now they've got you. This is one consequence of a larger design issue: QUIC is trying to provide "equivalent" security to TLS, but if your threat model is "something is wrong with TLS" that's not enough.

drewcrawford··on A Saudi Arabia Telecom's Surveillance Pitch (2013)
Shameless self-promotion, as this article is one of several that inspired me to work on this.

I have been secretly working for several years on a drop-in iOS technology based on the latest peer-reviewed crypto research, pinned certificates, PFS, and implemented in a safe language that doesn't buffer overflow.

It's invitation-only right now but many customers are running it in production. We have to position it as a performance technology (and it's very very fast) since security doesn't make sales. But my personal goal is to Make Things Dark Again™.

Unfortunately the hardest part of security work is funding it–performance and business concerns drive a lot of the conversations which limit how much time is left to work on making bulletproof transport security dead simple for app developers. And it needs to be dead simple to get deployed in enough apps to protect the average user from pervasive surveillance.

But if you believe in the problem, and your employer can create a budget for the problem, I would love to chat with you (email in my profile). There is still time to rewrite the future and keep totalitarian governments out of our communication. We just have to decide we actually care enough to make it happen.

drewcrawford··on Apple Fires Back at Spotify for Asking for 'Preferential Treatment'
It's way more complicated than that. Antitrust law doesn't care about "the market", it cares about "the relevant market", where "relevant" is a legal term of art that may or may not have anything to do actual markets.

There were 6 factors considered in establishing the relevant market in the Microsoft case (paraphrasing Court's Findings of Fact, §2)[0].

A) Servers were not cost-effective replacements for consumer hardware and were excluded from the relevant market. It's not clear whether the discrimination also goes the opposite way (e.g. instead of "smartphones" we compare "flagship smartphones"). If so, that kind of market definition would really make iOS look gigantic.

B) PowerPC and other non-Intel-compatibles were excluded from the relevant market since consumers could not install the competing OS on their hardware (that's right, "the relevant market" was constructed so as to exclude Apple, MS's only serious competitor). It is an interesting question whether the fact that you can't run Android on your iPhone for non-ISA reasons would be viewed similarly to an ISA difference in the hardware, but the way I read the ruling, it seems likely to be viewed similarly, or perhaps much worse. The reasoning all stems from whether consumers have choices to install their own OS, nothing to do with ISAs per se.

C) Various "borderline formfactors" were excluded from the relevant market. In the 90s this was e.g. WebTV, thin clients and game consoles. Today this might be things like tablets, VR platforms, chromebooks, etc. These are (today) all Android form factors, and excluding them to focus on "traditional" formfactors makes iOS look a bit larger than it does in Netcraft surveys.

D) The court observed that market conditions create a feedback loop, e.g. more people write apps for the dominant platform, that make the dominant platform more attractive to write apps for, round and round it goes, etc. This is obviously true for the smartphone market.

E) The court was concerned that nobody could realistically enter the market and "in less than a few years, present a significant percentage of consumers with a viable alternative to incumbents." This seems probable for the smartphone market, based on Mozilla and Canonical's misadventures, RIM and Nokia's inability to make a comeback, and even Windows Phone.

F) There was also a discussion about why "portals" (like today's cloud computing and/or webapps, basically) are not an effective substitute for doing computing locally and should not be in the relevant market, as well as a discussion of why Java is not a substitute for Windows (obvious to us, but not to lawyers)

All of this to say, it is really not so obvious from a legal POV why Apple are not a monopoly or are in any way different fro the Microsoft case. The real reason is probably nobody wants to sue them; I have seen estimates of the Microsoft case that suggests MS spent more than $500M defending it, and the taxpayers (with much cheaper lawyers) spent more or less $100M to prosecute it and didn't get much for their money.

Then there is the question of who gains by spending that money: Apple is well-loved by their customers so it doesn't have the election appeal for politicians that the Microsoft case did. Google would prefer to implement Apple-like policies in their store and don't want them found to be illegal. That leaves bit players like Canonical or Mozilla, and for them the legal costs are likely prohibitive.

[0] https://www.justice.gov/atr/us-v-microsoft-courts-findings-f...

drewcrawford··on Ethereum Protocol Developer Holds $115k Worth of DAO Tokens
Suppose for a minute the miners and the software developers disagreed strongly on some important decision. What do you think would happen?

* Would the software developers suck it up and follow the miners?

* Would the pool of software developers shrink because various developers do not find implementing the will of the miners to be an enjoyable hobby and/or job anymore?

* Would the miners hire new software developers to replace the ones that leave?

* Would there be a fork?

I don't know, but pretending that "the protocol is decentralised, the technology is completely agnostic to anything but the will of the majority" is ignoring the realpolitik of this kind of situation.

Saying that Ethereum implements the will of the majority is like saying American democracy implements the will of the majority. Sort of? But meanwhile there are a lot of empirical studies [0] demonstrating that the will of the people has little to no influence over American public policy. A democratic financial system is much more corruptible than a democratic political system.

[0] https://scholar.princeton.edu/sites/default/files/mgilens/fi...

drewcrawford··on An audio engineer explains NPR’s signature sound
> Is this not a blatant layer violation?

This is the one of the areas a recording engineer's problems are different than a software engineer's intuition.

In software, we have a lot of problems that can ultimately be traced back to not having good "visibility" into the software or its state. This has various far-reaching consequences, from specialized tools to examine software (such as debuggers) to cultural values ("one job per component") that help us reason about complex systems we cannot see.

But in the case of audio, we have a multiple sensory organs to directly perceive sound. A bad audio engineer would be able to push a few buttons to move his headphones around in the signal chain and know in seconds where the EQ is occurring. A good one would simply know the U87 rolloff "by sound" and would not have to conduct any investigation at all.

There are some cases where more detailed analysis is done with fancy tools–mastering song recordings, for example. But in live sound, an audio engineer's primary concern is UX, because good UX is the difference between solving a problem quickly and annoying pops or worse. One button that engages a good EQ is much, much better than something like this: http://adn.harmanpro.com/product_attachments/product_attachm.... And epoxying the single button to the correct position is even better still!

drewcrawford··on Maintainers Matter: The case against upstream packaging
Chromium may have been packaged in 2010, but it was released three years before. The Debian weak keys "cockup" may have been created in 2006, but it was discovered two years later. Maintainers had long windows of time to add value. Did they?

I don't mean anything personal by it. If I was maintaining 38 packages on my nights and weekends I'd do a bad job too.

But examples don't go out of date unless you present some force that takes 30k unmaintainable packages and turns them into 50k maintainable packages. What specific advance in software maintenance do you believe improved the art of software maintenance by an OOM?

But if you want to talk recent examples, we could talk about how nobody's packaging Swift.

drewcrawford··on Maintainers Matter: The case against upstream packaging
I feel like this article is sort of attacking a straw man.

> The promise: Sandboxing makes you immune to bad ISVs. The reality: It protects you from some things and not others.

Well yeah, but the same could be said of maintainers. Maintainers let things through all the time, and sometimes they cause problems (hello, Debian weak keys).

The reality is a bit complicated, and it boils down to something boring like: if the maintainers are better than upstream, maintainers are good, and if they are worse than upstream, then they are bad. But that is both tautological and vacuous, so it is not especially useful insight.

The real insight is this: there just aren't enough maintainers to go around. Debian has 1500 contributors and 57,000 packages, which is 38x as many packages as people. Now maintainers have a lot of tooling to make that more tractable, and upstream developers have their time split between multiple packages too, and some packages are important enough to get the attention of several maintainers fulltime. But do we really believe 1/38th of a person is judiciously considering how to overrule upstream's decisions, attentively investigating user bug reports, and so on?

More likely, the maintainers do not even claim to be doing that, because they have not packaged the software at all. I use lots of unpackaged software; until recently nobody was packaging Chromium for example. Java packaging is really unreliable. And so on.

Ultimately, Ubuntu Snap exists because there is a lot of unpackaged software, and a lot of software packaged poorly. It would be nice if we could wave a magic wand and get 38x more maintainers, but we cannot.

← PreviousPage 2 of 29Next →