Hacker News Reader

Comment by wereHamster | Hacker News Reader

  > 8. Open Source Is Becoming a Monoculture
  >
  > [insert comparisons how one software is vastly preferred over another,
  > eg. git vs subversion]

If you'd make the comparison five years ago, half of the software would not be present. svn would rule over cvs, abiword would rule over openoffice, a konqueror would rule over mozilla etc.

It may be a monoculture today, but nothing stops people from inventing the next SCM tool, the next word processor or the next web browser. That has happened in the past, that will happen in the future.

View on HN

This was the only one I disagreed with. Everyone said GCC was it forever, yet here we are ushering in an age of llvm/clang/etc. SVN was decidedly dethroned by git (and also mercurial). I think you can cherry-pick places where monocultures have developed but history shows that those are the places that get disrupted.

Anybody who said GCC was forever doesn't know their history very well.

See: http://en.wikipedia.org/wiki/GNU_Compiler_Collection#EGCS_fo...

I agree. I don't necessarily see diversity as a virtue, but more the ability to adapt. When choosing a tool, people will naturally pick the most popular one, provided it suits their needs. Open source allows for everyone to get up and move if a better alternative appears. It may be a 'monoculture' of sorts, but that doesn't have the same consequence as the biological parallel. Open source can adapt.

Diversity is a virtue in that (much like in natural systems) it makes the community as a whole less vulnerable to any particular problem. Consider, specifically, software viruses/worms for instance - heard immunity logic should fit just as well as in the biological case; monoculture leads to more epidemics.

It is not with a cost, though - moving between different options and trying to get some measure of interoperation can be hard, which is why you get some concentrations at any given time.

The key differences between natural systems and computer systems is that vulnerabilities don't kill a piece of software, and there's no interbreeding. You don't get two types of emacs, one of which has a buffer overflow, breeding to produce a new emacs without the bug. You get top-down design, where the 'many eyes' theory dictates that new vulnerabilities should be found faster in a more popular code-base.

Evolutionary theory is useless for describing software vulnerabilities (presently), because the software on your computer doesn't get to pass on it's genes, hacked or not. It's a clone of every other download from the same source, and they're all systematically replaced by a new release with minimal feedback.

I was speaking of epidemiology, not evolution. I agree that there are quite significant differences between evolution of software and evolution of organisms, and evolution certainly has an impact on epidemiology, but monocultures are an epidemiological problem for reasons mostly unrelated to evolution.

Agreed. In software there are costs to using niche tools, so typically everybody tries to use the most popular choices available that fit their problem space. This is true of all software, closed or open.

FOSS is always going to be more 'monocultural' in terms of what software is used. There are fewer costs to switching, and you get enormous benefits in FOSS from using a tool that everyone else uses.

Reply on news.ycombinator.com