Hacker News Reader

Comment by danso | Hacker News Reader

This is a pretty good example of how certain metrics aren't always relevant to reality, or, at least match the headline. The word "feed" implies that HN depends on the contributions from/links to these sites, but most users of HN would argue that domains such as github.com, github.io, nytimes.com, etc. are far more prevalent and important to HN than virtually any of the domains listed here. HN depends on daily, traffic...It's not that the sites with high medians aren't good, but they don't "feed" HN...Median score in this context is a trivial metric. Number of top stories, daily, by domain would be far more relevant in showing what "feeds the beast", as they say in the media business.

View on HN

domains such as github.com, github.io, nytimes.com, etc. are far more prevalent and important to HN

NYTimes.com is certainly more prevalent but, just like Grauniad lately, it seems to be diluting quality on HN rather than being important. Same goes for Medium mentioned above.

Whereas sites like patio11's or cperciva's blogs, YC startups (bu.mp), tutorials are what makes HN unique and interesting.

> it seems to be diluting quality on HN rather than being important

I think this underscores the difficulty in quantifying the nature of "quality", especially for a broad audience. I generally check the NYT homepage every day, so seeing its URLs on HN isn't particularly helpful to me (ignoring the value of the HN discussions)...however, there is so much interesting information on a daily basis, period, that I bet if the HN front page consisted solely of the most-upvoted of high-traffic mainstream sites, e.g. github, nytimes, medium...it'd still be interesting to me because there'd be a lot that I would've missed otherwise.

That said, it'd be cool to have an option/Chrome plugin to filter the frontpage links to domains with relatively rare submissions, just to be able to quickly see the unique upvoted submissions for the day.

I think this is what you mean: https://docs.google.com/spreadsheets/d/1mmfbNWaX0Nr1P65VmwZp...

[insert quip about techcrunch.com]

A little off-topic, but this reminds me of how non-straightforward it is to categorize what the true "domain" of a given URL. Blogspot.com has the most submissions by domain, but the API that runs the ?site query on HN returns just 5 links:

https://news.ycombinator.com/from?site=blogspot.com

I suspect that's because HN differentiates between somerandomdude.blogspot.com and googleblog.blogspot.com...but that in itself is an editorial/arbitrary decision...Why is subdomain more relevant for differentiation than for github, in which github.com/blog is grouped in with github.com/somedudesrepo?

And FWIW, it seems subdomain faceting is done manually...education.github.com links are shown on HN's page as just github.com...whereas googleblog.blogspot.com's domain is fully listed.

There is a well-defined solution to this problem: The Public Suffix List.

https://publicsuffix.org/

blogspot.com is in it. github.com is not, but github.io (where Github Pages are hosted) is. I would guess this is what HN uses.

That said, it of course can't help with e.g. categorizing github repo links by user, since those are by path rather than subdomain. Ah well.

I feel like the best way to deal with this would be to group by subdomain if a domain has more than say 5 or 10 different subdomains submitted, but group by domain if there are less. That way "hosted" domains subdomains like *.blogspot.com get their own category, but domains controlled by a single entity are treated like a single category.

Reply on news.ycombinator.com