6 ms·
"it's fiddly front end stuff I don't like all that much as a domain" What (briefly) is your experience or skillset so I can better target my recommendations?
by nickpsecurity 11y ago
"it's fiddly front end stuff I don't like all that much as a domain"
What (briefly) is your experience or skillset so I can better target my recommendations?
"A lot of what you suggest for directly solving it strike me as hacks not to my taste, but I'll look harder at that sort of thing in the future."
The problem is that the Web, Windows, and UNIX stacks are all hacks built on hacks with tons of compromises and many bad design choices. Anything that preserves their functionality without a rewrite is going to be hackish. So, we then decide where to put hacks among libraries, source-to-source translators, libraries, OS's, hardware, etc. I tried to name off more sensible one's to avoid readers having headaches I had going through 10,000 papers of this shit. ;)
"If you're going to modularize it, the interfaces are going to be complex and error prone"
Interface errors account for vast majority. Margaret Hamilton, Dijsktra, Robert Schell, and Paul Karger all independently discovered that w/ solutions presented. The consensus is you need to document the pre-conditions, post-conditions, and properties that always hold. Type-checking, Design-by-Contract, and similar techs help. Generic functions and good design helps keep complexity down. Functional languages with strong typing are best at making this safe and concise at same time. So, it doesn't have to be bad unless your tools suck.
"much more fragile cryptosystems/cryptographic protocols "
Sort of. Those do much more than just decompose: theory, design, implementation, distributed of all that... a tall order. Understanding the positive effect of decomposition and layering will be obvious if you look at this on p8 "Layered Design:"
http://www.cse.psu.edu/~trj1/cse543-f06/papers/vax_vmm.pdf http://www.cse.psu.edu/~trj1/cse543-f06/papers/vax_vmm.pdf
Now, try to imagine a description of how VMware or KVM do the same thing. It would probably be much more complex with looping around and such. The decomposition makes it easier to understand w/ sensible interfaces. That's how it's supposed to be done anyway.
"trying to design an public key based authentication server; no success then"
As I said, crypto protocols are extra hard to do right. It's the kind of thing that should get intense tool support. The state-of-the-art in high assurance is NSA and Rockwell-Collins tool that goes from a flowchart & specs in (Simula?) to generated code or hardware with verification checks for a processor whose security is verified. An example for the rest of us and ideal for your case is combining CRYPTOL for algorithm, state machines for overall code, and SPARK for implementation language. The protocol logic would be implemented with this kind of tool:
http://www.dsi.unive.it/~modesti/anbx/ http://www.dsi.unive.it/~modesti/anbx/
I call out plenty of people for their methods. Yet, it was understandably hard to pull off public key servers a few years after they were invented on 1970's computers. Even the people inventing INFOSEC sucked at security back then. ;)
"It's got to be fast; for now, ever faster as web pages get more and more complicated."
It does. Hence our hacks to work with normal, highly-optimized software. However, some old stuff was still faster. JavaScript is utter garbage. One alternative (Juice) used Oberon language for applets with ZIP'd abstract syntax trees. Let you type-check it, interface-check it, and then compile it super-fast to fast binary. Client-server apps could use efficient stacks, implementation languages, avoid rendering they didn't need, etc. So, plenty room for improvement especially if we use plugins or stand-alone apps where possible. Facebook Messenger, etc are ditching the Web that way.
"browser distribution that thoroughly pretends to be from a mobile source and gets those simpler pages. :-)"
Good thinking: I thought of that one, too. Still in backlog. Another model I worked out was ultra-fast, thin-client (phone or netbook) connected to a secure, desktop browser running on best hardware it could. Turns out that a variant of that (for efficiency) was done by mobile companies like Opera. So, it's possible.
"Objectively, I hear they're dramatically losing market share AKA $$$ for further development."
"that they're preparing to move to a multiprocess architecture which will nuke their extension ecosystem"
That is potentially bad. I don't know about their management. Can only hope for the best there. Good news is many of the best extensions are actively maintained or should be ported in event plugin architecture changes. So, here's hoping there too. This from Wikipedia bothers me:
"making Yahoo! Search the default search experience for Firefox in North America effective December 2014.[7] However, it is also to be noted that Yahoo! Search is a merely a front-end to Microsoft's search engine Bing, and Microsoft takes a 12% cut of all revenue from the search business per the deal."
Yahoo itself is in a shaky position and Bing gets a cut? That's risky esp with what they made on Google. I'd have gotten expenses in order instead. Maybe.
"starting with an Raspberry Pi seL4 port (I know, stop laughing, it's the most stable (long life promise)"
It and Beagle port. No laughing here: do what you can. Check out Genode as well as they're among more mature of L4-based architectures. Maybe better integrate it with seL4, Linux virtualization with either, or middleware for integrating components in protection domains. These are low-hanging fruit.
"Thanks a lot for your contributions to all these HN security discussions!"
Appreciate it. The few who might benefit are the only reason I post. Well, I enjoy it too but still. ;)
- hga 11y agoExperience/skillsets: "SCIENCE!", but because $$$ ended up doing systems and software, mostly backend with the occasional GUI, C, C++ (have sworn off of), Perl (ditto), the various Lisps, esp. functional programming, going to look at ML next. A bit of kernel, embedded, or akin experience, e.g. a MS-DOS TSR filesystem. UNIX(TM) since 1978. Also see https://www.linkedin.com/in/haroldancell https://www.linkedin.com/in/haroldancell Multics, ITS and the Lisp Machine are my ne plus ultra systems, but UNIX(TM) was good and available. I like the ITS/Lisp Machine development style, and in theory the FONC/VPRI and seL4 ones, but believe a modern web browser is a non-negotiable for users, and impractical with any of them, hence my angst. Retired by disability, but should still be up for a few more serious projects, hence my looking at this arena which addresses many of my pain points. Anyway: "JavaScript is utter garbage." Well, it has a lot of Scheme in it, so I'm not quite so hostile ... but I haven't use it except for a few months in 1997 ^_^. With any luck WebAssembly (https://github.com/WebAssembly https://github.com/WebAssembly) will make things a little less painful, the flip side of the "oligopoly" is only having to convince 4 organizations (Apple, Google, Microsoft, Mozilla). To restate, the problem for the foreseeable future is the 2nd part of Postel's law, "be liberal in what you accept from others"; an objective of mine is a "desktop" that doesn't suck and has survival value, which requires a web browser that accepts, shall we say, the "best effort" of others' web pages. Or why since the early days browsers would so their best to display something in HTML without closing tags. I believe "survival value" is very important, it's a quality of ultimate computer viri like the UNIX family that Richard Gabriel was getting at in his Worse is Better essays (https://www.dreamsongs.com/WorseIsBetter.html https://www.dreamsongs.com/WorseIsBetter.html). A browser's ultra-complicated moving target requires creating, co-opting, or allying with a big community of developers. Hmmm, and only by doing the latter are you likely to develop fundamentally better approaches, which need buy-in from enough of the browser "oligopoly". "Good news is many of the best [Firefox] extensions are actively maintained or should be ported in event plugin architecture changes." That's just not going to happen per https://news.ycombinator.com/item?id=10097630 https://news.ycombinator.com/item?id=10097630 and https://news.ycombinator.com/item?id=10099240 https://news.ycombinator.com/item?id=10099240 although I suppose there's time for Mozilla to change course, but that would require their current management to get a clue or be replaced. One bottom line for me is that Servo should be a Maximum Effort instead of a "research project". There's a lot more to be said, but probably best by email if you're interested (hga@ancell-ent.com). Me-> "starting with an Raspberry Pi seL4 port (I know, stop laughing, it's the most stable (long life promise)" "It and Beagle port." I seriously considered the Beagle ecosystem, but noted it's not "stable", TI makes no representation about continuing manufacture of any given board, and the earlier (and $$$) non-Bone ones don't seem to be available. But the BeagleBone Black (BBB) was a surprise smash hit (good price + very capable industrial controller SoC, e.g. can implement a 100MHz logic analyzer), so much so it had limited availability for more than a year, and now two companies are making them. But the Raspberry Pi huge, and committed to continued manufacture of their boards, albeit with respins. And a lot cheaper, BBB $55, first gen Pis $25 and $30, new Pi 2 $40; for just messing with seL4 I bought a $25 A+. The TCB is ridiculous, albeit ideal for for OEMs, but all things considered I judged it the best for an initial learning platform. If the BBB shows staying power I'll seriously consider it.