Schneier on Security

Clive Robinson • April 23, 2016 4:31 PM

@ Bruce,

Just to add to your workload 😉

It appears the UK’s Home Secretary Theresa May is steadily loosing contact with the reality of the economic side of National Security,

Buried in the draft of the IPA “Snoopers Charter” she is foolishly championing, is a requirment that companies ensure that the UK Gov is fully cognizant of any product long prior to it being launched,

http://www.zdnet.com/article/uk-spy-bill-will-force-tech-firms-to-disclose-future-products-before-launch/

Depending on how you read it the implication is if you design/manufacture/market a product within the UK you have to fully clear any product with the UK Gov… Or they will give you jail time, fines, seizure of property etc…

It has become quite clear she has not got a clue what the legislation means or does, and that she has not been properly briefed on it (otherwise she would know the simple explanation of what “data about data” means). All of which suggests that neither she or the civil servants in the UK Home Office are actually the generators of the legislation. Which begs the question of who is and how they have such a significant influence on what is extrodinarily dangerous legislation.

Clive Robinson • April 25, 2016 4:01 AM

@ The usual suspects,

This TRNG might be of interest,

http://altusmetrum.org/ChaosKey

For those not upto speed on things the CCT of the noise source on the page has two parts, the left hand part is a switch mode up converter turning USB rail voltage to around 20V which then powers via a resistor the two transistor noise source.

Whilst it will work, it does not consider noise from the switcher, or other ground noise.

But for a quick and dirty home brew which will work as a prototype it’s a good place to start.

Clive Robinson • April 25, 2016 6:57 AM

@ Wael,

So many posts and questions…

By the way, what does CCT stand for?

It’s shorthand for CirCuiT as well you should know by now, if you’ve associated with European engineers.

The designer is a highly respected fellow

I’m not sure but I think I met him at an AMSAT do a few decades back. AMSAT ground station was once at Guildford University Surrey UK run by Martin Sweeting –now Prof Sir–, the Uni kind of branched out to become Surrey Satellite Technology (STL) in the Uni business park, and developed all sorts of interesting payloads including CubeSat stuff. The Uni then floged their share to EADS and I’ve not had contact in quite a while.

If you can still excercise, then you’re not that old.

Trust me you can be old and excercise a lot if your body will alow (which mine often does not).

Any way “age is relative”, if you are the oldest in any group then you are “the old XXX” to the whipersnapers.

@ r,

Now you see what I have to putup with from Wael, so your tiny drop in comparison to his ocean is OK “So N’worries Mate”.

As for Wael, I intend to find out when he is next going to be in Death Valley, lying on his back stargazing, and sneak up and give him one of my evil laughs, and see if his hair turns whiter than Methuselah’s from shock, or if he irrigates the place 😉

Clive Robinson • April 25, 2016 8:15 AM

@ Thoth,

Maybe Linux developers should really consider switching from /dev/random to Fortuna RNG created by our host,

Two things, firstly if I remember rightly Bruce co-designed it with Neils Ferguson who went on to work at MS and had trouble with an Elephant.

Secondly, the real problem is not the algorithm but the source of entropy and how it arives at the instance of the algorithm.

Old Hard Drives being mostly mechanical had quite a bit of entropy from the drive speed and seek times etc on the heads. As tolerances became tighter, and read ahead caches in the megabyte and up range got put on them the entropy became less and less. Now we have SSDs the entropy from them is tiny and variations in times are more likely to be algorithmic issues not mechanical so effectively determanistic.

The user tapping at the keyboard or playing with the mouse, used to be considered a source of entropy but in these days of virtual terminals to headless servers, that one goes out the window as well.

As for using interupts as a source of entropy that can be quite dangerous depending on how the interupt occurs. That is it can be sufficiently determanistic that the entropy is not gatherable.

Then there is the issue of Virtual Machines, they don’t realy have the sort of hardware access that produces entropy. Worse if they do get entropy via the base/host OS then the chances are all VMs will get the same entropy which means various attacks become possible.

Thus it’s not just a Linux problem it’s a more general problem which hits all OS’s and all VMs, and is only going to get worse as we further design entropy out of hardware designs.

As @Who? Points out with,

Do they really trust the CPU RNG being truly random?

It’s a question of faith over commonsense. I’m on record here and other places as saying I don’t think the OnChip TRNG’s are anything of the sort, especialy Intel’s (google my name and “magic pixie dust”). The funny thing is Linus made a bit of a blunder over using such RNGs and he kind of had to do a facepalm out of it.

Aside from the usual suspect bugs, getting RNGs wrong has been a serious security failing since before the WWW existed… I’ve designed Hardware TRNG’s off and on since the early 1980’s and I’ve got a feel of just how difficult it can be to get them right.

As @Wael questioned my categorizing the noise source in the CCT, I will make a few observations about it.

Firstly the switching regulator is by no means properly noise decoupled. The single cap shown is in no way sufficient, and has a poor frequency range, thus will alow both low frequency and high frequency switcher noise onto the 20 volt rail. Further the transistors are connected to ground thus as they are current operating the variable ground current caused by it’s effective impedence back to the USB power source is going to be adding a lot of ground noise.

The designer mentions problems with frequency rolloff thus having to use a 200Mhz GBP OpAmp… But still uses a frequency selective coupling cap that will look like an inductor from a few tens of MHz. Further if you are going to use an OpAmp it’s adventageous to use it in a differential manner.

So I would firstly split the load resistor into two resistors of half the value and put one from the supply to the transistors and the other from the transistors to the supply return. I would then use three capacitors such as 1uF, 1nF and 10pF to replace the decoupling cap and likewise on the transistor pair outputs. I would issolate the supply with a balanced filter with a fairly low frequency rolloff. I would also take a differential output from the the top and bottom of the transistor pair and put it into the OpAmp configured for differential operation.

In TRNGs I’ve designed in the past I actually use two OpAmps the first used differentialy with very low gain and DC coupling to the transistors and gain increasing with frequency, this then fed a second OpAmp AC coupled with sufficient gain. The frequency selectivity was designed to give as flat a frequency response as possible in the bandpass upto the digitizer. Most homebrew digitizers are either a slicer, Schmitt trigger or over driven OpAmp into a digital input. Needless to say they have their disadvantages. These days I would use a highspeed 12bit A-D converter. But in the past I used to take the analogue output of the amplified noise source and use it to drive a Voltage Controled Oscillator, the output of which was used to drive the CLK input of a D-Type latch with the Data input driven by another high frequency oscillator. The raw output of the latch drove a shift register and a counter which I described on this blog a short while ago.

Clive Robinson • April 25, 2016 12:14 PM

@ Thoth,

How useful would the PUF function of a chip be for contributing to entropy ?

No more than any “hidden” –supposedly– unique serial number.

@ Nick P,

On other end, I’m trying to create a list of the benefits of HW TRNG over a software CRNG.

It might be easier to produce a list of weaknesses for both first then see which offsets the other, and what remains, that then needs addressing.

I suspect that when the final analysis is done, it will be an amalgamation of the entropy sources of a TRNG and CS algorithm or two to provide an entropy stired CS whirlpool, and a per process/VM whitening proces prior to a CS hash.

Clive Robinson • April 26, 2016 2:05 AM

@ Albert,

Is it possible to design a true random number generator that fits in one tiny package?

I see no reason in the laws of physics why it can not be done. Though there are good reasons why you might not want to use some methods (it’s not that dificult to convert a smoke detector for instance but do you want Polonium or similar on your work bench or in landfill?). Most good sources of entropy age in some way due to entropy it’s self, which means that they have a recalibration or end of use time. As a very rough rule of thumb, the easier it is to get entropy out of a source the faster it is going to wind down under the influance of entropy thus the shorter those times will be.

Thus we tend to use entropy sources at the bottom of the curve such as thermal noise, which means two things in general,

1, The source is susceptable to interferance.
2, The amount of entropy in bits/Sec is very low.

Which gives rise to a real design issue.

The entropy signal is consequently hidden in other often much larger signals normaly.

So what you expect to see from a source output in descending order of magnitude are,

1, Recognisable determanistic signals.
2, False entropy, determanistic signals.
3, True entropy.

Whilst removing the recognisable determanistic signals is possible by generating an inverse signal and subtracting it or by other filtering techniques, it tends to be a resource intensive task.

But that does not solve the real problem which is the false entropy, which could best be described as “A determanistic signal you can not determin”. The only way you can limit some of it is by very carefull circuit and physical design to isolate the source from influence external to it. It tends to be a resource intensive thus expensive and physicaly large solution.

Which brings us to your second point,

It might be easier to get MB manufacturers to include one in their designs.

Which is realy a basic economic question of how much of a demand for good entropy sources do MB purchasers have? And the answer is “almost none”, thus after removing any profit it gives you the amount of money the MB manufacturers will spend on it, which is considerably less than they would spend on a power supply filtering capacitor.

IC manufactures put them into their devices because they are essentialy “free space fillers”, thus in effect cost them nothing whilst giving a small marketing advantage.

Thus the general MB manufacturer will spend nothing on putting a TRNG onto their boards.

There are a few other economic arguments, but the reality is the probability is such that COTS equipment will now never have TRNGs that you can trust implicitly.

And as has already been pointed out on this blog, the non general, specialised or niche product manufactures are now in the UK and US LEO cross hairs with legislation already in draft form to make it all nice and legal… Oh and the legislation appears to have no limits as to what constitutes a manufacturer, thus “home hobbyists beware, your cards are being marked”.

Clive Robinson • April 26, 2016 6:59 AM

@ Anura,

Well, the other day I realized you could make the key schedule so it is its own inverse

The earliest known example of that is probably the “reciprocal cipher” given in the Kamasutra under the art of “Mlecchita Vikalpa”.

Perhaps more commonly known is that of the Enigma Cipher machine used by the Germans and others during WWII where the reflector stator at the left of the rotors made it a reciprocal cipher. Such rotor machines became the basis of many designs (like sigba) that remained in service untill the microcomputer caught up in the early 1980’s. However others were aware of the problems with reciprical ciphers and did not have a reflector plate.

The problem is that in many ciphers where “the key schedule… is its own inverse” there is the problem of “no input may encrypt to it’s self”. Unlike the Polish Bomba[1] which was a ciphertext only attack on the key indicator, the British designed bombe was designed to make use of assumed plaintext that used the “no input may encipher to it’s self” to find a jibe with the cipher text the coresponding letter to letter corespondence was used to set the menues being used. Gordon Welchman further realised that the reciprocal weakness could be further exploited which gave rise to the “diagonal board” that caused Alan Turing to stand there mouth dropped when Gordon explained it to him.

Thus you have to be very careful that you don’t fall into the “no input may encipher to it’s self” trap when designing key schedule’s that are their own inverse…

[1] In the early part of the war the Germans did not use the Enigma in a sensible way. One such weakness was how they sent the individual “message key” under the “Day key” twice at the begining of each message. The Polish unlike the British had guessed the internal keyboard to rota cage entry wiring of the Enigma. This enabled them to use the six charecter key indicator to break into the actual day setting. Around Oct1938 Henryk Zygalski, invented a way to speed the process up by using a stack of perforayed card “grills” on a light table. These became known as “Zygalski sheets” but were manpower intensive to use. Thus the Poles automated the process and it was this device that was known as the Bomba. Unfortunatly the Germans changed the way they sent the message key, and the Polish Bomba became obsolete. It was this obsolescence issue that nagged at Alan Turing and he tried to design it out along with Doc Keene in the British Bombe design, which made it a whole different animal in nature, especialy when Gordon Welchman independently came up with the idea of the diagonal board.

Due to the security policy put in place in the 1980’s by Margaret Thatcher, the actual truth about the bomba/bombe diffrences did not come out for a number of years (till around the fall of the Iron Curtain). It was not helped by the “Official History” by Harry Hinsley nor by other historians who had no actual contact with either the Polish or British cryptanalysis of the Enigma.

Clive Robinson • April 26, 2016 3:45 PM

@ Bruce,

I assume this is more amusing to people who know about My Little Pony.

I’ve been told to tell you… That the “A Word” should be “Adorable” not “Amusing”…

Apparently it’s something that “real men” should know having “communened with their inner feminine child” or some such crystal waving new age thing…

Clive Robinson • April 27, 2016 3:46 AM

@ Figureitout,

At those speeds, one little freakout and spinout and your car is rolling and you go flying…

I don’t know if you know or not but research findings have changed.

Back in the fourties and fifties you have a small knock and your car would wrap around you and spill gas every where, and the death rate was consequently very high and your chance of dying rose closely related with speed. Well things have changed a bit and safety features in cars make the risk of dying not related to speed but how the impact energy gets distributed.

Modern designs of tank tend not to spill gas as easily, and one traffic officer I’ve spoken to has said that they now treat vehicals catching fire and killing people as suspicious untill shown otherwise these days.

But the odd one is risk of death on a motorbike. The way the figures stack up you are actually less likely to die at speeds above forty five MPH than you are below them with the risks rising again only after around one hundred MPH, if you wear the right clothing.

Why the figures are as they are takes more analysis to tease out. And it appears that your chance of dying depends more on not what you hit but who hits you and from which direction.

For instance it you get struck on the driver side door at as little as thirty your odds are realy not very good. Worse if you are a passenger in the back and your door gets the impact. But even if you are sitting on the other side of the car, the human body is not realy designed to take sideways impact, and that standard seatbelt is likely to do you more harm than good as the first point of restraint would be your neck.

The upshot as one car designer I know involved with race cars, put it, “You can ride in comfort and die in pain, or you can ride in discomfort and live to an old age”. That is your risk of death and injury is now dictated more by how you are confined during a crash, than the design of the car outside the passenger compartment.

Which is problematical, because there is no such thing as “Mr Average”, race drivers tend to have custom seats designed around them. It was a lesson various Air Forces around the world had to learn about fighter pilots and one those who travel in cars should learn, if they want to survive impacts from unexpected directions with only bumps and scratches.

Clive Robinson • April 27, 2016 6:18 PM

@ Miles,

Former Tor developer created malware for the FBI

Why does this not suprise me in the slightest…

The lack of hard information gives me one problem though, which is trying to decide if it should be classified as “an insider attack” or not.

Presumably at some point someone will take a detailed look at his work and decide if he added vulnerabilities or not, and if he did, did he later exploit them. If he did then it does not bode well for how people view his ethics, or for his future employment (would you trust him enough to employ him, afterall he could be “backdooring” you?..).

Clive Robinson • April 28, 2016 5:41 AM

@ Figureitout,

Why not a circular buffer here? If a circular buffer gets corrupted a stack would too?

As far as coruption is concerned, all memory bits are as likely to be corrupted. However the glib answer is, circular buffers have two pointers which means twice the chance of pointer corruption than a stack with only one.

Which hides the point that stacks actually have three pointers, stack bottom, stack max and current stack top, and circular buffers four. The difference is that in both cases two of the pointers the start and end position in memory can be hard coded into ROM etc.

However it’s the way the two behave in terms of implementing error detection and correction that makes the real difference.

In hostile environments you need to not just ED&C on a memory location basis –which works much the same for both– but also on the entire memory range as well. It turns out this is much easier to do on stacks and is only reliant on the hard coded menory start and end pointers and one additional independent read pointer, and importantly the bulk of the checking can be asynchronous to the working pointer, thus can be carried out by an ibdependent process running on a different CPU or hypervisor built in hardware.

A simple example of checking would be two dimensional parity checking. You can do a fast vertical 16 or more memory word block just by incrementing a pointer and XORing the word at that address into a register. Then only if that fails[1] do you need to do the slow parrity checking on individual words. This is easy if the blocks are effectivly static as they would be in a large stack where the only changes are at the stack top. This would not be true of a circular buffer where blocks get overwritten more often, and their position change each time the pointers wrap around.

There is also a more interesting aspect… Memory gets effected in several ways, firstly on the static side by random probability, and by adjacency to a bit that has flipped (clustering). Secondly the more subtle dynamic effects, where the various logic gates set to read or write a bit make bit flipping in the rows and columns more likely (reminiscent of rowhammer as I’ve mentioned before 😉

The upshot is in a stack you need only check blocks closer to the head of the stack or in binary relationship to the head of stack address on a frequent basis, older blocks are best left alone, and checked only occasionaly.

There have actually been specialised books and research papers written about implementing ED&C in comms and storage, but I’m not realy aware of those in English (I have a list of Polish ones somewhere as the PhD who did the thoretical side got her quals in Poland where in the 80’s the cutting edge practical research in EC&D in mainframe memory was being done, as I neither read or speak Polish I’ve not got copies nor can I tell you which is most appropriate).

[1] This is a simple example which assumes only one bit flip in a block. However the principle still works on more complicated error checking.

Clive Robinson • April 28, 2016 7:51 AM

@ Dirk Praet,

This might be of interest to you,

http://uk.businessinsider.com/saleh-abdeslams-lawyer-calls-him-stupid-2016-4

To quote,

Saleh Abdeslam… is a “little moron from Molenbeek, more a follower than a leader.”

According to Saleh’s lawyer, Sven Mary. Other comments are even less complementry.

Whether it is a pre-trial ploy or not is difficult to decide.

What ever the reason, Saleh hardly sounds capable of tying his shoe laces unaided, let alone use “sophisticated covert communications” that some drum banging politicos have previously claimed (no real suprise there ;-).

Clive Robinson • April 28, 2016 3:28 PM

@ Bruce,

Not sure if you still have students to challenge with the Fermi Paradox, but… It appears that no evidence stacks up for Type III (galactic engineering scale) civilizations.

http://arxiv.org/abs/1604.07844

@ For those who want to know more about the Kardashev scale,

https://en.m.wikipedia.org/wiki/Kardashev_scale

On the Carl Sagan variation of the Power Kardashev Scale we are not yet a Type I being a little under 0.8 on what is a log scale, so we have a ways to go yet. It would almost certainly involve space based collection, but the real technical problem in not the collection but transmission of energy over the required distances.

The reality is that it would probably be easier to utilise energy at the point of collection and move industry into space, which is why NASA amongst others is keen on developing the “Space Elevator” concept. Especially when the consideration that the ultimate form of polution is heat is taken into account. Thus radiating the heat from inefficient engineering / manufacturing processes would also be easier in space than from earth back into space.

Clive Robinson • April 28, 2016 7:15 PM

@ ALL

If the facts presented in this article are correct, then the US is not the place to be.

A police sargent has been jailed effectivly indefinitely under the “All Warents Assist”.

The sargent has not been charged with a crime nor has he any previous charges.

The Magistrate accepted something that was effectivly false testimony by a forensic investigator that “his best guess” was that an encrypted drive contained child pornography (something he can not know if the files are encrypted, and guessing is not knowing and is thus not even hearsay let alone qualified opinion).

The authorities have decrypted another drive, but have found nothing incriminating.

Thus the only other supposed “evidence” is the man’s sister who has claimed she went over to his house and watched child abuse images with him (which is a very odd thing for her to say as it would appear she made no attempt to contact the authorities over it at the time… Which begs the question was she compelled to say it by the authorities?).

http://arstechnica.co.uk/tech-policy/2016/04/child-porn-suspect-jailed-for-7-months-for-refusing-to-decrypt-hard-drives/

The EFF has submitted an opinion to the court that “compelled decryption” is testimony, and thus in breach of the fifth.

It’s now over due that the likes of congress should be reviewing the use of the ancient “All Writs” legislation.

Clive Robinson • April 28, 2016 7:55 PM

@ Nick P,

Yet, your post did make me wonder why two of you said to run from ROM instead of RAM.

First things first, the design of ROM (not FLASH EEROM EPROM etc) is quite different to that of RAM. As a rough rule of thumb you can consider SRAM to use either the equivalent of a latch, or capacitor on the gate of a FET as a “stotage element” that can have it’s state changed by design or a high energy particle. In the case of “fuseable link” or “mask programed” ROM there is no storage element or sense amplifier that can have it’s state changed.

That is not to say that ionizing particles can not damage a ROM but you are talking a very significantly larger energy required that is actually blowing holes in the semiconductors.

Further I actually said run an interpreter that is stack based using three stacks. Effectivly you use the stacks as a voting protocol to replace just one stack. The interpreter I ran actually had nine stacks aranged as voting tripples, one tripple for data another triple as a pointer stack and the last tripple could be used as a program stack if required (think modified Forth Interpreter). Aside from the interpreter, there was code run via an interupt, that did actually checksum the ROM code, as well as brown out detection and other sanity checks.

Some of it was overkill but when they wheel it into a vacuum chamber and start firing radiation at it you want the best chance it will pass the –very expensive– tests. Think of it as EMC susceptibility testing on steroids.

It was actually talking about this to someone about what they could do for their degree project a decade or so later that gave me one of those “gut reactions”. It kind of said to me it was the way to go on security… Which in turn kicked off the thinking that led to the basis of the Prison idea some time later.

And they say “Good ideas never happen in a vacuum” 😉

Clive Robinson • April 28, 2016 8:42 PM

@ Nick P,

Just had a quick read through the link you gave.

Most of it is what I have a habit of doing, there are a few other tricks as well.

However the bit about external oscilators was at best wooly, and Im guessing not written by a person who has actually “done the nasty” in terms of design.

As a rough rule of thumb there are two types of Xtal you will come across HF and above “AT Cut” and LF and below “watch crystals” usually at 32KHz that run the “real time clock” these are an entirely different cut and are VERY fragile.

The trick is to use the on chip inverter as an RC oscilator at 32KHz and shunt the resistor with the watch crystal in series resonance mode. You need to pick the resistor(s) with care to ensure that the voltage swing and drive power etc stays well within manufactures specs, and the capacitors provide the correct loading for the crystal with the chip capacitence accounted for. This often means you end up with something like a 2.2 Meg Ohm resistor divided into two values in series and the crystal shunts the resistor going into the inverter input, with that capacitor being used for frequency trimming.

When the oscilator starts up, it will not see the crystal so it starts oscilating at the RC frequency, this will get close enough that the crystal will go into the low impedence series mode oscilation and thus take over. this usually works fine, however if you are “Doing a Marvin” you can make a bridge circuit and look for the change in voltage level as the crystal oscillates.

There is an old electronic design engineer saying about the problems you used to encounter, which is “Oscillators don’t and amplifiers do…” (oscillate 🙁

If you can find a copy back in 1973 Motorola had the blue “MaCMOS” data book, it had a whole load of information about the care and feeding of CMOS gates, including how to get inverter gates to “act as amplifiers” you could then use to make your RC oscillator and crystal oscilator.

Clive Robinson • April 29, 2016 5:41 AM

@ Figureitout,

There’s some pretty elaborate dynamic algorithms…

There are many such, but many are but angels dancing on a house of cards. You need to look at them with the assumption that they are well above the CPU microcode level in the computing stack. And it does not take long to find out that Intel and others can change the Microcode to RTL interpreter on your CPU at will… And in Intel’s case make it a requirment that you do so at boot time… Hence I don’t design using any Intel products these days and don’t advise any one to do so either. That is an across the board ban, even where they have spun business units off.

The important point is that, you have to look at “pretty elaborate dynamic algorithms” with a great deal of skepticism and ask “how do I protect against bubbling up attacks?”. Most algorithms it’s way to difficult if not impossible to do in software alone. Some however you can see how to augment them such that you can get the security down to just the likes of the Instruction Pointer register for single point of change attacks, others to even multiple change points of attack.

With out going into messy details you use tree based state machines that lack feedback or feedforward paths, voting protocols and various types of efficient check sums.

So simplisticaly you have the “hidden IP” that walks through your state machine code. Each time you update the state and it’s variables, you immediately checksum the whole state in a way it can be easily be verified. As you call subs you pass the state checksum, the state and the data that caused the sub to be called into the sub so it can check if it is being called validly and entered correctly. In this way you can to a certain extent tie down most things including the instruction pointer.

If you also write the state machine so it “fails safe” when an error does occure you can restart the state machine from a safe point.

It does make for one heck of a lot more code, and very carefull thought and design to check you always have the bases covered in a verifiable way but… You can in effect build a simple interpreter that runs on what is a virtual “paranoid CPU” with very limited instruction set, that is also self verifying.

You will not catch everything with just the virtual paranoid CPU but you can implement basic hypervisor type activities in the interupt system of the real CPU as well as have external state checking hardware.

The interesting thing is though that when you have taken the time and effort to do it the first time, you almost always want to carry it forward into other projects, kind of like a “comfort blanket” 😉

When I did it for my first 16bit micro (68K) in assembler, I implemented a striped down version of Forth (yes I know you hate it) and later “Small C” which was a reduced K&R compiler back then.

Have a look at that link Nick P gave, you will find it and some of the places it points to of thought provoking interest.

Clive Robinson • April 29, 2016 8:49 AM

Is this where US&UK Crypto Laws are taking us?

The EFF have investigated Arab Nation “Speech Crime” where laws are only used to chase down those providing on line comment the authoritarian leaders dislike.

As the EFF this is a case of identify the writter then having decided on the punishment, apply a law to achive the punishment. Thus “rule by law, not rule of law”,

https://www.eff.org/deeplinks/2016/04/crime-speech

It takes only a moments thought to realise that proposed US and UK crypto law that mandates access to any and all communications without warrant or legitimate judicial oversight is a first necessary step towards, such Speech Crime…

Clive Robinson • May 6, 2016 11:19 AM

@ Dirk Praet, Nick P and others,

Perl tends to be a bit of a maintenance nightmare if not properly documented…

There is a joke that “Perl is a very WORN language” on the assumption you know WORN is an acronym for “Write Once Read Never”.

Like many a SysAdmin that had to look after a mixture of OSs the fact that Perl was available and ran almost the same way on all of them was something I was thankfull for. But… the much claimed “strengths” of what it could do, I found out weighed by the fact that it’s syntax etc was not intuitive and easily got wrong, thus you had to have a “blue camal” on your desk as well as an appreciation for warped humor…

And just as you got used to it in version 4/5 they anounced six that was “So far off the Larry” you started thinking “BASIC is looking good today”.

I must admit I’m many another programers nightmare, I often find it’s quicker to code up from just the language basic operators rather than go hunting through doc after doc of poorly documented libraries.

It has much to do with my engineering background where you have to “carve” your own electronics to real world transducers many of which can rip you to shreds and spit you out as juice teeth and all. Then design the fatal to non fatal level voltage interface which connects to what most programers would consider to low level to be seen hardware like analogue, and status IO lines on microcontrolers. Then write a low level BIOS / RTOS and cut code in assembler for device drivers, and “up convert” to USB or 10baseT etc. All so that those other code cutting programers can moan, you’ve not given them enough to work with because it does not work with Widget-lib5000 etc etc 😉

I guess it’s one of the reasons I have real downers on CS grads, where as physics, engineering, chem and various lower end biology grads who have had to earn their spurs the hard way, with learning how to get their real world experiments to deliver data to a PC, as just the first step, I tend to get on with…

Clive Robinson • May 6, 2016 12:05 PM

@ Nick P,

Hmm,

I’ve already described on this blog how to do a big chunk of this.

However there are some things that cannot be reliably done.

Clive Robinson • May 6, 2016 4:04 PM

@ Nick P,

One of reasons I posted it. Put in a proposal. Make you some money and build a secure app. 🙂

I’m not a US citizen and as far as I am aware DARPA do not hand research grants to non US non resident people/organisations.

However $150K for a year barely pays for 1.25 man/years. As for the $1mill for two years that would just about get you two people for two years plus office space somewhere cheap oh and a bit of kit…

Anyway I’m probably not the sort of person who would get past their grants/appropriation committee. Because the last time I was aware it was checked I was on the US IC “shit list” over the design of electronics of military significance destined for the “Axis of Evil” back last century. I was actually supprised they even let me in the country a few years ago back when they had just set up those shite awful finger print systems at imigration. Maybe the UK IC had finaly sorted it out with them, maybe not, either way I’ve no intention of going anywhere to find out they don’t have an orange jump suit to fit me, the last thing I need at my time of life is chafing of the crotch. And unlike those of Matrix-Churchill I was working on rather more than a nod and a handshake, read the Scott Report if you want to know what a clusterFcUK it all was at the time under Maggie Thatchers guidence. The plus side was that the then independent “Customs and Excise” got the first substantive nail in their coffin on the three strikes and out principle they lost their right to prosecute and got subsiquently subsumed into their hated rivals the Inland Revenue.