Ok, I'll dare: per the mythical man month, the biggest factor limiting the success of programming teams is communication. A demographicly uniform team can communicate more effectively with each other (due to having a larger shared vocabulary of common experiences) and this reduced communication overhead greatly increases their productivity.
>If there were something keeping women out of the community, you'd want to fix it. Right?
This metrics vs. data thing is precisely why hearing people go on about the lack of women in programming makes me nervous. It reminds me of my manager going on about how our unit test coverage is only 60% and why isn't it 70%; if you define low test coverage as the problem rather than the symptom, we could easily write some terrible tests that would increase the coverage but make our codebase worse.
So likewise, if you start thinking about "only 4% of open source developers are women" as a problem /in itself/, it's very easy to jump from there to doing something that increases the number of women in open source, while actually making things worse (e.g. "let's offer a load of bursaries for female CS undergrads to work on open source projects"; quick, easy, increases that percentage, but could quite possibly backfire by giving women in open source a reputation for low-quality code - and more to the point, does absolutely nothing to address whatever actual problem you're referring to)
Aren't you arguing against OSS here, essentially? Is OSS not the perfect example of communication across cultural and demographic differences? Should Torvalds just shut out everyone under 40 in the name of efficiency?
It reminds me of my manager going on about how our unit test coverage is only 60%.
Splitting your time between coding and test coverage is an extremely poor analogy to inclusivity in software. For one, your time is a resource, software is output. If you split 8 (wo)man hours into 8 hours testing and 0 hours coding, your output goes down; if you add 8 (wo)man hours to a software project, your output goes up. Even more importantly, coding and testing are not people, they don't have feelings, children to feed, etc. The comparison simply doesn't apply.
As far as your point about "bursaries" (had to look it up, it means scholarships), please consider that the original article talks about having more women's user groups and better conference policies and makes no references to financial aid of any kind. It's really not germane to pit every discussion about gender in software against apocalyptic scenarios. Yes, it would be bad if all of a sudden men were totally excluded and women were totally included, but we are far from that now, aren't we? Shall we worry about "loads of bursaries" when said loads become an actual problem?
Of course not, that would be putting the cart before the horse. He should recruit those who work together most efficiently - but if it turns out the most efficient team is demographicly homogeneous (and I suspect that the core linux team is) - then why not? Are you going to argue that he should form a less efficient team in the name of diversity?
>Splitting your time between coding and test coverage is an extremely poor analogy to inclusivity in software. For one, your time is a resource, software is output. If you split 8 (wo)man hours into 8 hours testing and 0 hours coding, your output goes down; if you add 8 (wo)man hours to a software project, your output goes up. Even more importantly, coding and testing are not people, they don't have feelings, children to feed, etc. The comparison simply doesn't apply.
You've misunderstood; my point is not that by focussing on testing you take away from coding. My point is that, while a low test coverage number indicates a problem, if you just focus on the coverage number you can end up doing more harm to the problem you were trying to fix.
The post I replied to was saying "look, if only 4% of OSS developers are female then clearly that indicates some problem." My point is that if we focus too much on the number, then we're only going to treat the symptom. Let's fix the problems, but let's not put too much attention on one shallow number.
Edit. You say: the post I replied to was saying "look, if only 4% of OSS developers are female then clearly that indicates some problem." I've re-read the post that you replied to, and it neither said, nor implied anything of the sort.
Why?
Because its a moral judgment, we can assume that even if monocultures are functionally superior with regards to performance we still will prefer a diverse culture for the intangible benefit of 'good ethics'.
No, wait, it occurs to me that "some Indians" would require an India, and therefore a kind of monoculture there to start with. Perhaps a "diverse culture" would mean there is no India--there would be equal proportions of all conceivable backgrounds and stuff everywhere, or at least something close to that.
But, now, wouldn't that society be a monoculture? A very serious monoculture, with the vulnerabilities that go with it (i.e. lacking the defenses that come with diversity). To give an example: if every city looks the same and has the same number of each kind of building and uses the same proportions of transportation methods and so on, then if a virus gains a mutation that makes it go very well through a particular kind of transportation, it'll be able to infect every city equally well. The situation goes from "virus no one cares about" to "virus able to infect everything". On the other hand, if you had rich cities in which one type of transportation dominated, poor cities in which another one dominated, some cities that were hubs of international travel and others that rarely got any visitors from more than 15 miles away--then you'd go from "virus no one cares about" to "some cities overrun, others barely affected", which makes it much easier to leisurely prepare a response with aid from the barely-affected cities.
(Genetic diversity is good for fighting disease, for similar reasons.)
Mmm, I have reached the conclusion that the goals of diversity at the level of individuals found in a group, and diversity at the level of groups found throughout civilization [you could define any number of group sizes if you wanted], cannot be simultaneously achieved to their fullest extent. If you have no monocultures, that puts some limits on how diverse your set of cultures can be. This meta-problem gives me a kind of zen feeling.
Well, if you have it, demographic uniformity is a good thing indeed. If anything it translates to less racial tension, more intra-community trust etc.
That doesn't mean that it's good to expel people, minorities etc, to achieve it if you don't have it. But if you do have it it's translates to fewer problems along the way. In the end, you either struggle FOR integration (ie towards "demographic uniformity") or different groups struggle to exclude the others (ie towards "demographic uniformity" again, but in their favor).
Just ask any African country that has been plagued by different ethnicities and cultures into war for power / control...
"Lack of diversity" means that people are excluded from a community, either actively or passively. That's a "bad thing" pretty much by definition. You're thinking about the metric ("how many women?"), while the author is thinking about the mechnism implied ("something is keeping women out").
The former is data, and I guess you can argue that it's neither bad nor good. But surely you agree that the latter is bad, right? If there were something keeping women out of the community, you'd want to fix it. Right?