HNHacker News
TopNewBestAskShowJobs

bmir-alum-007

246 karma · joined July 10, 2015

submissionscomments
bmir-alum-007··on At 78, Scientist Is Starting a Hedge Fund
http://web.archive.org/web/20150724000231/http://www.wsj.com...
bmir-alum-007··on The Weakness of the .NET OSS Ecosystem
I've worked on open+closed source in enterprise startups with people that come from both MSFT- and Linux-centric worlds respectively. The MSFT centric-people tend (not all) poo-poo open source as a complete waste of time and/or lack of business savviness (helps competitors too much). There is some open source contributions in such shops, but usually only if it helps (and doesn't seem to hurt) the bottom-line.

PS: Unfortunately, most Fortune 2000 enterprise (ms and non-ms shops) abuse the fuck out of open source and give zero back (time, $ or help). I really think there should be some "community charge" funding model which doesn't make it nakedly for-profit but rewards people that contribute, maintain and support useful code that is widely deployed so they're not completely abused as unpaid crowdslaves.

bmir-alum-007··on YC Backs Portable Coffee Stand Called Wheelys (YC S15) to Take on Starbucks
I worked at Stanford in the dept which supported housing and dining. Concessions agreements and relationships have to be engaged and signed before setting up shop in an area, regardless of whether someone believes it to be "public." In this instance, it would have meant engaging smctd. [0] It seems like they don't have any sort of formal concessions process or outreach, which is business-unfriendly (uphill battle). Otherwise, the relationship will be adversarial / nonexistent and the business will be finished unless it can legally skirt on a technicality (by moving just far enough out of the muni authority's jurisdiction). There are certain municipalities / area which need certain types of permission... CalTrain land is not technically under the local city's jurisdiction... They like other U.S. transportation authorities have their own police and own rules, like Stanford (ZIP code 94305) and some other universities too.

0. http://www.smctd.com/

bmir-alum-007··on Cassette Revolution: 1980s Tape Tech Is Still Making Noise in Our Digital World
Yeap, "Weird Al" Yankovic and all gathered dust after CDs became readily available. I had a $400 Sony Walkman CD player (circa 1986) and a 1981 (!) German impression of The Doors - The Doors album (dull orange silkscreen printing) with the aluminum reflector material extending all the way to the edge of the disc.
bmir-alum-007··on Wikileaks Global Intelligence File Dump Contained Malicious Software
Virustotal doesn't run locally, there's zero client impact!
bmir-alum-007··on Pragmatic physically based rendering: Intro
IIRC from reimplenting 1/2 of the OpenGL pipeline in C for a CS course, the early Phong and Gouraud shadings inevitably looked either plastic, matte or dull. Hair and skin were hard but have been mostly solved fairly okay at great computational expense. And there's lots of hacks for things like oranges (bump mapping -> surface normal tweaks) and then they're shader languages. Each surface material (wood (especially chatoyance), textured marble, dull cloth, broken cement, butterfly wings (structural color/schemochromes), conchoidal broken glass, irregular dirt, dead leaves, pearlescent automotive paint, oil slicks (thin-film interference), on and on) needs slightly and/or completely different hacks to pull off the "Hollywood" real-time smoke-and-mirrors, least CPU/GPU/memory intensive fakery to render a believable scene whereas actual Hollywood can throw massive resources offline, at their leisure to render a scene frame-by-frame.

Metal is extremely difficult to model because light travels oddly along crystalline and micro-fissure structures.

bmir-alum-007··on Show HN: Send free postcards worldwide
Ah cool.

Also, what data is collected permanently, for how long? What happens if the venture where acquired? Whom would own that?

bmir-alum-007··on First Java Zero-Day Attack in Two Years Targets NATO and US Defense Organizations
It was written clean-room in some client's office (no NDA, but still, engineering ethics), but I may have a copy somewhere on a backup in offline IRL storage, but it was written for JDK 1.2-1.4 IIRC so it won't run and it's chocked full of confidential / proprietary apps (which it installs). It called a few C Win32 APIs to create icons and write registry entries, on Linux I think it just wrote files and maybe ran `ldconfig`. It was super dumb, fugly but it shipped and worked consistently. There was even Swing (pre-AWT) skinnable branding background images. It was probably 450 lines of Java and 80 lines of C. Autoplay even worked decently on Windows and it came up in under 8 seconds because I optimized the ISO image for file load order to minimize seeks, and that's booting an entire 1.4 JRE from a 4x CD-ROM drive to a usable, interactive state.
bmir-alum-007··on Wikileaks Global Intelligence File Dump Contained Malicious Software
Exactly, hence the preference for a partnership directly to get some more API access priority / infrastructure $ for dedicated scanner worker bots (virustotal is probably a bunch of internal VMs with COM instrumentation agents to WinRM or CLI Linux scanner versions, so more of those for private/dedicated users * ).

* I think virustotal could be easily monetized (without altering or cheapening itself) by freemium to orgs, nonprofits, govts that want faster, private, shared/dedicated access to the same "cloud" capabilities. There was HitmanPro which was good early on, but it didn't seem to be stay anywhere near as comprehensive as virustotal.

bmir-alum-007··on Wikileaks Global Intelligence File Dump Contained Malicious Software
Please have a look at what I wrote... Not tamper, inform (least surprises).
bmir-alum-007··on Bringing the best of Bash scripting to Ruby
It's a tradeoff. There's also nothing stopping folks from rewriting init and rc scripts as Ruby code or (re)writing systemd in Go, Ruby, etc.
bmir-alum-007··on Bringing the best of Bash scripting to Ruby
Ruby scripts ultimately are processes, and there's much borrowed from Perl including `ARGF`. Just don't use the clever Ruby syntax stuff unless one wishes to be bashed <groan 2.0> over the head.

It becomes more of a "what should this program should consume, do and output and feedback" problem... Bash, Ruby, whatever... but aim for some standards, in general.

Furthermore, with Ruby, as a dynamic language with full network capabilities and add-on packages (gems) with other IPC libraries (zmq), there are other structured, efficient serialized service consumer/producer configurations possible rather than only shell commands. There are also background job processing frameworks like sidekiq and parallel job processors like celluloid, parallel and EventMachine.

The future, really (for dev &| ops tasks) in on-demand loaded, micro-services rather than giant, slow frameworks. The glue (say chef code) and one-ofs will likely still be a bunch of Ruby code and hopefully more Ruby code, respectively, too.

bmir-alum-007··on Bringing the best of Bash scripting to Ruby
The script I use to disable/enable dnscrypt-proxy (multiple instances behind dnsmasq for HA) https://gist.github.com/steakknife/bf6967e2298b95da29b7

Sample output (tethering):

    0 ~ git:master ❯❯❯ internet_access                                                                            
    Bluetooth DUN:  Bluetooth-Modem  dns 127.0.0.1
    Ethernet:  en0  macaddr xx:xx:xx:...:xx  dns 127.0.0.1
    FireWire:  fw0  dns 127.0.0.1
    Wi-Fi:  en1  macaddr xx:xx:xx:...:xx  dns 127.0.0.1
    iPhone USB:  en4  172.20.10.3/24  router 172.20.10.2  macaddr xx:xx:xx:...:xx dns 127.0.0.1
    Bluetooth PAN:  en3  dns 127.0.0.1
    Thunderbolt Bridge:  bridge0  dns 127.0.0.1
    0 ~ git:master ❯❯❯
Sample output (wifi enable, unassociated, no tethering):

    0 ~ git:master ❯❯❯ internet_access                                                                              
    1 ~ git:master ❯❯❯
This is driven by a non-privileged infinite looping event bash scripts which TCP connects to another pair of bash scripts running as root which can actually make the change because they're listening on two localhost tcp ports.

(I've tried these other event-based app/script utilities and they're fragile and buggy as heck.)

It seems like something Apple, for one, should take on... deprecate AppleScript and use something more common, accessible and make a common, standardized scripting platform far more comprehensively documented in something like Ruby. AppleScript is just too COBOL/Cucumber "English" and poorly documented to be practical for advanced users. (I've been gradually moving all new scripts over to Ruby because Bash scripts are fragile and not as inherently reusable as standardizing on a dynamic languages (like RedHat w/ Python).)

bmir-alum-007··on Wikileaks Global Intelligence File Dump Contained Malicious Software
Edit: One box had really crazy, clever malware (backdoor IRC bot which was firmly detectable by NIDS (snort) and by remote nmap) which defeated local nmap, portmon and rootkitrevealer... and it was a 24x7 production oracle box (no HA or archive log mode master-slave repl) running the dining order, meals and inventory databases, so live cd / usb-hdd-specialized hdd dongle (better forensics) weren't possible.

0. If I had budget authority, I would've ordered something which could grab memory and disk images on a live system and sign-up multiple .edu researchers & symantec security group and equivalent shops under NDA to analyze them.

1. And I would've yanked all those janky (R)ILO (aka RMSA, aka DRAC) cards with their always outdated Linux / Java / PHP "wifi router"-like whatever embedded systems.

2. Finally, I would've spent some cash on honeynet setups and cc: to item 0.

Edit^2: Props to Josh Wieder for taking sounding the sec awareness alarm. I would only do active sec research on untrusted materials within a decent hypervisor's VM on a virtual desktop (VDI) which has "nonpersistence" on all storage, so it's clean on every power cycle.

bmir-alum-007··on Wikileaks Global Intelligence File Dump Contained Malicious Software
It seems like a scraper which could manage submitting to and checking virustotal would be a good idea.

Also, perhaps wikileaks / virustotal could come to sort of agreement to scan things and flag them while still making assets available to researchers whom still need raw docs, but with large virus warnings which intercept downloads to a warning landing page confirmation if there's no referrer (direct, deep linking).

Finally, the other issue is that not all malware (past, current or future) is known. We ran Windows Servers (NT, 2003) boxes directly on the internet (before I forced them to offer departmental firewalls) and these boxes had already attracted all sorts of multi-vendor unseen malware, rootkits and backdoors long before I got there. After deploying WOLF to a number of boxes, Mark Russinovich was like "yea, you should probably take [the IT sec depts advice]" by just get firewalls and then format all these boxes back to clean image, fully-patched and strong, production-usable, security policy state. Most IT folks don't conduct forensics research... they see strange behavior, try to find a scanner/remover (maybe 80% success), or partially remove it and continue despite unprovable to fully remove all traces... Or they wipe the boxes, start over and guess/pray that the same 'sploit doesn't exist twice. There's just too many unknown variants of existing sploits and too many new sploits to have much faith in antivirus (it's reactive, last line-of-defense security). Antivirus is an opposite Bloom filter... it'll tell you if you're pwned or may not pwned. It's a good to have, just not a complete holistic security posture.

bmir-alum-007··on First Java Zero-Day Attack in Two Years Targets NATO and US Defense Organizations
Yeap. It's preferred to vendor a JRE (non-system JRE) inside your app (which is basically just extracting it to some dir you always control), which doesn't conflict with whatever else is on the box (no registry settings, /etc/profile or common locations like /usr/some/unix/path/to/system/jre ), and auto-update your app which includes updating the JRE. This is how to deploy a supportable app, which reduce support tickets about "you broke our apps, our IT department is going to call our VPs and is going to call your VPs" and pre-sales issues like "we can't buy this because we can't change JRE versions."

https://www.java.com/en/download/faq/distribution.xml#eula

bmir-alum-007··on First Java Zero-Day Attack in Two Years Targets NATO and US Defense Organizations
Most JVM methods can require permissions because it provides built-in POLA ACLs (security policies). The default would likely be to deny network access and file io. It's more likely that an app itself contains vulnerabilities &| unnecessary weaknesses due to ease of misconfiguration because of the pervasive lack of full-stack implications, grokable and auditable security policies.
bmir-alum-007··on The Disadvantages of Single Page Applications (2014)
In general, it depends: there are many different use cases where SPA is fine and others where they are not. It's especially good for long-running consoles that don't need much navigation (given using the appropriate JS framework which doesn't leak resources on most browsers). SPA doesn't make much sense (apart from in-web app search live results searching) for static pages that need to be SEO crawlable and can be served up fast from static CDNs. Perhaps the core engineering tradeoff is whether to breaking the stateless RESTful contract by managing state with session cookies and/or JavaScript, and doing so in a way that may or may not be sharable to other users, history-friendly, SEO-friendly, etc.

Ruby's turbolinks is a clever hack and mostly works approach to make nav-based (browser history- and crawler SEO-compatible) almost as responsive as SPA... it replaces assets, title, meta and body on-the-fly and updates the url location with javascript to save a whole page update. It's basically a hybrid of client+server coordinated fragment caching. There are some gotchas and workarounds for onLoad() and other JS hooks, but it mostly works pretty well.

bmir-alum-007··on First Java Zero-Day Attack in Two Years Targets NATO and US Defense Organizations
JVM & CLR runtimes probably account for 75% of all enterprise backend apps running on server gear in production right now. Both have been beaten on for so long that they're both fairly reliable, which is not to say either bug free (reliable or secure), resource efficient or maximally productive, just more usable in production on average.

On the client side (browser-land), JavaScript (and HTML & CSS) is ubiquitous. The issue of course is that every browser is "especially" different. So it makes no sense to deploy browser extensions unless things need to talk to other local apps. It makes more sense to develop hybrid native apps which are mostly html5 using something like QtWebKit and add your own per-platform goop rather than try to use something like node-webkit, phonegap (cordova) or appcelerator, because, inevitably, system-specific code will be needed, and not having direct access to APIs because of limited proxy/thunk/trampoline code doesn't make for happy developers. Java tried to do the hybrid way using JNI, which could load dll's, dylib's and so's dynamically (I wrote a custom, cross-platform (Linux, Windows), CD-ROM-based app installer using Java + JNI in 1995 by the way, it was trivial.. And then I wrote a Java source to MIPS binary, direct, non-JIT compiler in C++ in 1999.).

bmir-alum-007··on Show HN: Send free postcards worldwide
De facto nonprofit.
bmir-alum-007··on Show HN: Send free postcards worldwide
That's not the issue. Is the uncertainty of unwittingly signing yourself or the recipient up for junkmail or yourself for more spam.
bmir-alum-007··on A 2.7 mile long straightedge
In case anyone were wondering, gravitational lensing would only account for 1.6030569323305754e-13 m or 0.16 picometers of laser deflection when the laser would be tangent to earth and the ruler is extended from that same point.

Deflection approximated by:

    tan( G * mE / (r * c(air)^2) ) / 4345 [m]
bmir-alum-007··on The myths of bioinformatics software
Disclaimer: I used to work at a Stanford bioinformatics shop.

There's a clear need of AWS-like features for bio/biomedical informatics specifically enabling sharing, security, reuse and anonymization of data (PHI), libraries (like R's bioconductor) and infrastructure (IaaS/PaaS/SaaS).

The issue is that some labs archive still archive their data on actual hard drives (USB and bare drives), making their data much less useful than somewhere readily available and sharable.

I think it's a huge (billion+) opportunity where the right execution would need loads of smart, consultingish customer service reps (huge overhead costs) to help researchers with coding, sysadmining and bio to some degree. Basically, a full-service (with self-service, a-la carte features) hosting company for bio / medical.

This space is only going to grow deeper and wider as more is discovered and confirmed about each gene, protein, pathway and each accompanying expansion in nosology. This sort of research knowledge is vital and unlikely to shrink. The main issues are that it would be a cash-intensive and undefensible business model because it requires paying lots of consultant/scientist brains and anyone can copy the model.

← PreviousPage 4 of 4