Noah's ark.

Have you ever actually tried building this 18GB SVN tree?

It's certainly possible after you're forced to be deeply familiar with how it works (and write few custom scripts/makefiles), but obfuscating the process on purpose to prevent forks shows no good faith. This is borderline hostility towards GPL even if they try to be technically compliant. OEM firmware vendors are "compliant" by posting half-assed .rar of some outdated dev tree, dare to ask me to sign NDA (seriously broadcom? what the fuck with those kernel patches) and calling it a day.

I was just trying to add my language translation of the webui and ended up soldering serial to debrick my device few times.

OpenWRT is what GPL should look like, free, super easy to mod and test out changes, bloated and not very user friendly.

Sad truth is DD-WRT offers best stock features (luci is certainly lacking in some areas), especially on 4M flash devices.

PS: Eventually ended up compiling custom /etc/web and /usr/bin/httpd and fmk'ing existing binary image. Both files are interlocked to "prevent rebranding on ebay" causing major pita. </soapbox>

They live strong, reincarnated in MLC NAND flash cells exactly because flash was first thing to reach litography cost limits. It's not actually true analog, but close enough to keep precision.

SHA256 double hash applications were probably first who used this on massive scale. It's actually ok to ramp clock/voltage up 50%, get 30% more rate at cost of 5% of wrong answers (and halving MTBF). ASIC miner chip giving wrong answers now and then because of imperfect mask process (even before OC) is common too.

However numbers for standard-cell ASIC design don't seem much favourable, certainly not "doubling", much less energy saving (on the contrary, at ballpark 10-30% of OC you reach point of diminishing returns, and only if you dont care much about MTBF).

Now what would be interesting is actual "analog" computers, ie number of states anywhere between 4-inf - there is literally too much of wasted "potential" nowadays. NAND flash chips do it already because they are about to hit limits of cost-effective litography (10nm?).

It is the distro with the best cutting-edge version

You had me almost sold there.

AIDS

obligatory superjail episode

Anything you can't do in GUI.
Agreed, both DD-WRT and OpenWRT GUI can pretty much do the same thing, and I did recommend DD-WRT to amateur users in the past as it appeared to be somewhat more consumer focused.

Like custom samba settings. Or just want to use stuff which is up to date, not 3 years old binaries. Or openvpn server with authentication against remote ldap database. Geeky stuff. Scary, scary command line :)

Optware stuff falls a bit short in that regard, mainly because I want the router functioning without unreliable external storage and it is basically just a hack - making custom DD-WRT is messy (FMK? really?). Openwrt supports baking images nicely from day 1 using buildroot.

When you are forever alone, everything is possible! Ganbatte!

I love MikroTik as much as nearly any other eastern european do - the gui/cli is just plain awesome and dumb proof - even people unfamiliar with networking are able to pick it up quickly, compared to the "horrors" of linux routers or cisco-cli.

Sadly, your point is somewhat moot - AR9132 chipset of RB2011UAS is just home ap SoC and that is it. Routing performance is generally god-awful to make any use of BGP or OSPF. People generally just buy high-end RBs or run routeros on PC to do any kind of ISP networking.

RouterOS comfort comes at a price too - it is just linux kernel inside, but 3 years outdated, undebuggable corner cases etc etc. It works ok most of the time, but only very brave people are running eBGP on RouterOS or any core backbone for that matter. Wiping low-end routerboard and installing openwrt there won't help much either - people do that all the time with the very same SoC chipset, from tplink, edimax etc plastic boxes which come 30% cheaper and are the same utter crap reliability-wise (usually PSUs).

OpenWRT on cheapo commodity hardware - personally I'm using TL-WR1043ND, 4x1gigE/300mbps 2.4ghz N, USB storage is best bang for 50 bucks.

The system is reasonably specced to run openvpn gateway for home network and serve USB drive miniNAS via smb.

DD-WRT is basically GUI polish for people who don't wan't to delve into scary command line, but otherwise nowhere near as flexible as openwrt is.

Let me tell you exactly what happened:

Some vendor quit the business for whatever reason. He then posed as a hacker of himself, providing "proof" of the hack (passwords, adresses etc), demanding 500k. He then proceeded to lay out the "I owe money to these drug people, thats why I need the 500k" story. He somehow provides DPR with "real life data" of this hacker.

Then he poses as the group the hacker owes money to. He then accepts the offer of the hit, photoshops some picture and collects 150k from DPR.

Occams razor smoking dope. Why would the group accept 150k if they were owed 500k? Why did the data turn out to be fake and why was there no murder or missing person filed in the area? How would they even carry out a hit within such short notice in far away Canada?

Bonus points: The 500k and the vendor were undercover shills, that's never gonna show up in court proceedings for tactical reasons (aka entrapment in criminal law).

DPR got soceng'd hard.

John Brunner, Shockwave Rider, 1975

Lucky northern EU bastards. Mid-east europe, well generally anywhere with telcos pricefixing data transfers (that includes U.S.), its like 1GB-5GB cap with usual 20EUR/30USD plan. Classic story - its simpler to milk users with their 2yr contracts and provide crap service in duo/triopoly cartel, rather than actually compete and develop decent 4G infrastructure. Mostly 3G coverage, 4G only parts of big cities. Weird exception is romania which is almost as good, or even better connected than .fi/.ee for some reason.

Yes, sometimes there are rumors circulating about what probable problems there will be, but never closer specifics. So yes you have to memoize and practice a lot of "useless" stuff to even stand a chance. Even if there will be offline version of wikipedia and wolfram available during those contests, memoizing at least the common stuff gives you incredible edge - because it's just not the knowledge but also an acquired skill of choosing and applying common algorithms to various problems.

Regarding "new" data set, I know how it feels, exact same thing (corner case data set) happened to me few times.

The key is to be not bitter and throw tantrums and baseless accusations the moment results are published, but actually look over the winner's code (heck, even post it online to wider forum, if you dare :), and find corner-case data set too, during the appeal period.

The first time I had to admit a loss - the guy's solution was perfect (or I was unable to spot the bug). The second time the competitor's code had similiar degree of a problem, just not so obvious from reading the code. Technically it would be tie, but good judge will give you bonus point for pointing out his mistake.

Now, this was in highschool state championship in small east-european country, not sure what ACM rules are, but I suspect it will be something similiarly reasonable

don't attribute to malice what can be explained by incompetence
Judges are human too.