Really big projects

Posting mode: Reply [Return]
Name
Email	#♨Y@)4⛤b☜P,.9☘z}^Zct{\1xw0~ HKa⚠B:r6$v5/7GnDX MyF'A\|<pRl(E=uj+iW♊OLQNh_m>e
Subject	Spoiler Image
Comment
Flag
File
Embed
Password	(For file deletion.)

File: 1608526428001-0.webm (4.89 MB, 360x386, marble-framework-size.webm)

File: 1608526428001-1.png (137.34 KB, 1276x4670, LOC-2015.png)

Really big projects Comrade 2020-12-21 (Mon) 04:53:48 No.6226

Seven hundred million lines of code written over a period of thirteen years. That's how large the CIA's Marble framework was at the time of its 2017 Wikileaks "Vault7" exposure. At Bill Binney's estimate of $50 per line, that's a $35 billion program.

Ever since learning this I have been attempting to find programs of a similar size. The Linux kernel is currently roughly 27.8 million LOC. Windows Vista was 50 million LOC. Mac OS X v10.4 was 85 million LOC. Supposedly some "high-end" proprietary car software is at the 100 million LOC range. And according to many sources, the entirety of Google internet services is 2 billion lines of codes–but I don't think that really counts because it's hardly a discrete piece of software.

So is the Marble framework one of the largest pieces of software ever created? Is there nothing else that compares?

Comrade 2020-12-21 (Mon) 04:53:48 No.6230

is LoC really sound metric to use for measuring size of a project? Language like C can really blow up when you attempt to replicate object oriented programming concepts (systemd code base is verbose not bloat IMO). I don't know anything about proprietary car software but if they are one of those real-time control things that needs to be written in assembly or very low level constructs, they will also blow up in terms of LoC.

if there is a way to measure information density of a program by complexity of feature and then combine it with number of features, that will be really cool application of information theory.

Comrade 2020-12-21 (Mon) 04:53:48 No.6231

>>6230
well OK then I'm just counting number of instructions of binary but there could be a way to measure this in more abstract ways through flow analysis or something

Unique IPs: 1