Comment by jcr | Hacker News Reader

Comment by jcr | Hacker News Reader

1.) You never set background-color. Not setting background-color results in unintended consequences on any browser configured to have a default other than the expected white background. A good example in your case is visited links become grey on grey (assuming a browser with grey default background).

2. You need to be cautious of and properly handle notorious webspam/blogspam regurgitation sites like "recode.net":

"Wal-Mart Scammed Into Selling PlayStation 4 for $90" http://recode.net/2014/11/19/wal-mart-scammed-into-selling-p...

Which leads to:

http://www.cnbc.com/id/102197050

Which leads to:

http://consumerist.com/2014/11/18/terrible-people-create-fak...

Which leads to:

http://kotaku.com/people-are-scamming-walmart-with-bogus-che...

Finding the original source can be a PITA, but it's still much better than rewarding spam regurgitation sites with traffic.

3.) Having a "sign in" icon link on every single article listing looks odd. Is it the default replacement for some voting mechanism that's only available to logged-in users?

4.) Grey text on a white background is common fare on the web (including hn) to de-emphasize some text, but it's a poor design decision in terms of accessibility. People with even minor visual impairments (i.e. poor vision) have difficulty reading grey text on white backgrounds.

All in all, it looks interesting. Good Luck With It!

1. Got it. We'll set a background color in the body.

2. I agree. We will think about ways to handle it. As a last resort we may blacklist those kind of publishers.

3. Sign in is only needed to favorite stories (using the little bookmark icon below each line). I agree that it is not clear what the sign in is for, we will fix that.

4. Yup, we will increase the contrast a little.

Thanks for the feedaback!

On #2, using binary block/allow lists for sites works, but long term, it can be problematic in the sense of maintenance effort (i.e. your time ;). A less common but more interesting approach is to use site-based weighting. There will still be some manual wrangling involved, but you might be able to automate some of it eventually (analysis of content on the site, dns, linked-from analysis, google page-rank lookup, etc.)

With site-based weighting, you can still block with a threshold, but you gain the benefit of adjusting rankings based on past/known site quality, and the advantage of setting with a "default weight" for unknown sites you've never seen before.

It's really just a different approach that you might want to consider to handle the webspam/blogspam problem.

We already have mechanisms like weighting, whitelisting, blacklisting, etc. in place. We generally trust a good quality source (site domain) to produce good quality content, but there are exceptions. Since our popularity scoring algorithm relies heavily on social signals, it trusts the crowd to reward good content, and ultimately good content will rank higher.

>notorious webspam/blogspam regurgitation sites like "recode.net"

Actually recode.net does a lot of original reporting in addition to regurgitation. Disclosure: some of my former CNET/CBS colleagues are reporters there.

The problem (and I've seen this when building http://recent.io/ as well) is that every news organization does this to some extent. If organization X has a scoop, Y and Z will "follow" it by summarizing and rewriting it. This has been going on for over a century; the AP, founded in the 1840s, does it very well.

In the most egregious cases like the one above it's just a theoretically fair-use excerpt from the original story followed by a "read more" link. But sometimes it serves your users to link to the followup coverage instead. That's when the original article (FT, WSJ, Economist, etc.) may be behind a paywall, or when the followups have more context or additional details.

TL;DR: This is a non-trivial problem.

Even Google News only mostly gets this right: it tries to group stories on the same topic, but sometimes it has multiple buckets for the same topic, and sometimes an unrelated story gets thrown into the wrong bucket.

2) And if you follow the links and you end up at a Gawker site, probably best to forget the whole thing.

Reply on news.ycombinator.com