How to talk about feed auto-discovery?

genmon commented 4 years ago

When feed auto-discovery works, it's a great user experience, and lots of newsreaders promote entering the website URL in the "Add a feed" box as a convenience function -- so it needs to be mentioned. (As otherwise users will be thinking "where do I enter the Feed URL?")

But I'm reluctant to recommend it as the primary route as it feels like an expert feature, so it doesn't belong in a Getting Started guide.

In my (informal!) testing, the experience of auto-discovery was:

Go to a site that looks like it should support RSS (i.e. it has updates, it's reverse chronological, it's structured like a blog)
Give the site URL to a newsreader (in this case, using the iOS Share sheet to hand it to NetNewsWire)
It didn't work for the first 2 sites we checked -- and when using the Share sheet, you don't even get a "Couldn't find a feed" alert

After inspecting the source, it turned out that these 2 sites didn't support RSS. So the learnings are

RSS is less well-supported than it was
For a new user, the experience of trying to use auto-discovery and it failing is just baffling

So in the first instance, that's why the site promotes a longer but guaranteed successful path.

But auto-discovery does need to come up... but how? Maybe as part of an "Advanced usage" guide?

philgyford commented 4 years ago

Just to turn things around... you could think of auto-discovery as "this is how feeds work", while the "Advanced usage" is having to find the Feed URL and copy-and-pasting that.

Because ideally every site/page that has feeds will have them mentioned in their <head> and just using the site/page URL in your feed reader will display all the feeds you could subscribe to. Simple! It's certainly the case for all/most WordPress sites (30% of the web). I can't remember the last time I wanted to subscribe to a feed and this wasn't the way it worked for me (maybe I'm only subscribing to well-coded sites...).

louispotok commented 4 years ago

I agree autodiscovery deserves mention, and I think this could slot in pretty easily to the bottom of section 3.

How about the following?

Sometimes it's possible to subscribe to a site's feed even if you can't find the feed link on their front page. To do this, copy the sites "regular" URL (like nytimes.com) and paste it into your newsreader just like you would with a feed URL. Your newsreader should tell you whether it can subscribe to the feed, or not. To learn more about how this works, see [here](TODO: find link).

I can't find a good link at the moment, though -- what I've found is mostly how-to guides for adding the functionality to your own site.

Also, I'm curious -- what else were you thinking might go in the Advanced Usage guide?

genmon commented 4 years ago

Because ideally every site/page that has feeds will have them mentioned in their <head> and just using the site/page URL in your feed reader will display all the feeds you could subscribe to. Simple! It's certainly the case for all/most WordPress sites (30% of the web). I can't remember the last time I wanted to subscribe to a feed and this wasn't the way it worked for me (maybe I'm only subscribing to well-coded sites...).

I may have had bad luck in my informal user testing, and that's colouring my view... So I'll go through my thinking, and maybe saying it out loud will help undo my biases!

The need for step (3) is instructions that work almost every time (regardless of site or newsreader), and in the case that there's an error, there are "repair" instructions that let the user recognise and remedy the error (again, regardless of site or newsreader.

Then to figure out what those instructions are:

The user has to look for evidence that there is a feed (like a link or an icon). There are sites like the Leith's blog that show all the signs of offering a hidden feed (it's called a blog, there are titles and dates and categories, etc) only there's no feed. Because this was the first site that came up in my informal tests, I assumed that it was unlikely this was a one-off. The experience of trying to subscribe and it not working is awful, because a new user won't know whether it's a problem with them following the instructions of not. Conclusion: whether subscribing with a Feed URL or autodiscovery, the user should only make the subscription attempt if they have already seen evidence that there is a feed.
Finding the relevant URL.
- In the case of the Feed URL, it's a right click/long press. This is a pretty painful process, I admit.
- In the case of auto-discovery, there are two things to describe. First is how to get to the homepage for the blog. This is necessary because if the user lands on e.g. a Marginal Revolution blog post, the auto-discovery URL points at a feed for the comments, which is different from the feed icon on the page which points to a blog feed. There's a wrinkle that the homepage for the site is not always the root of the blog. Second there are instructions for how to copy the URL in the address bar, which isn't as painful as right-clicking for the Feed URL, but it's still a select-all and copy
Subscribing
- In the case of the Feed URL, you subscribe and either it works or it doesn't
- In the case of auto-discovery, it will either work, or appear that it's worked but you've actually subscribed to the category feed or the post comments feed (so there need to be instructions to check for that case), or it doesn't work and you have to fall back to the Feed URL approach. I don't know what the numbers are around auto-discovery support (the relevant tag needs to be added manually when you're making a site with a static site generator, for example), but the bigger challenge is ensuring the instructions work for all newsreaders: many sites that use auto-discovery advertise multiple feeds. How do different readers handle multiple feeds, and how should the users be guided to choose the correct feed? This isn't necessarily a major challenge, but it does require doing research into the top newsreaders and checking the UI for each.

So that's the thought process that led me to using Feed URL instructions.

I definitely think auto-discovery has its place... but my feeling about how to use it is to have an interactive page on aboutfeeds.com where the user can paste in a URL, and it's more like a "wizard" process that can catch all of these edge cases.

Writing this out, I'm aware I may be being too paranoid! But good to have this logic in a post so I can remember what on earth I was thinking in the future...

genmon commented 4 years ago

Also, I'm curious -- what else were you thinking might go in the Advanced Usage guide?

To be honest, I'm unsure... How to move between different newsreaders (export/import) might help people feel more comfortable choosing that initial newsreader. A guide to discovering more feeds perhaps? Though that's a tough one.

louispotok commented 4 years ago

@genmon Appreciate the extremely thoughtful and detailed reply! Definitely helpful for me to understand your thinking and I suspect it will form the backbone of the content we end up needing here.

It sounds to me like we have different ideas about the acceptable level of failure. I read your last as being very focused on making sure at every stage the user does not take a wrong or ambiguous action. My thinking was that we could basically say "this section not guaranteed to work, but it might." I think there's a way to split the difference here and end up with a good synthesis.

My suggestion would be to add 1-2 sentences at the bottom of section 3, mentioning that it's sometimes possible to subscribe to a website's feed even if you don't see a feed URL, but it doesn't always work. Then link from there to a separate page laying out the ways it can go wrong, the signs to look for, and what to try as mitigation. That would be the place to put in all the caveats you laid out about comment feeds, category feeds, site homepage vs blog root, etc, plus workarounds for those.

What do you think? If that sounds good I can try to write it out.

philgyford commented 4 years ago

The user has to look for evidence that there is a feed (like a link or an icon). There are sites like the Leith's blog that show all the signs of offering a hidden feed (it's called a blog, there are titles and dates and categories, etc) only there's no feed. Because this was the first site that came up in my informal tests, I assumed that it was unlikely this was a one-off. The experience of trying to subscribe and it not working is awful, because a new user won't know whether it's a problem with them following the instructions of not. Conclusion: whether subscribing with a Feed URL or autodiscovery, the user should only make the subscription attempt if they have already seen evidence that there is a feed.

Conversely I've come across sites that show no visible evidence of having a feed and yet auto discovery shows that they do! (Of course, I can't think of an example now.) Maybe their template doesn't display the feeds, or they purposely got rid of that geeky stuff, but either ignored or purposely left that stuff in the head.

In the case of the Feed URL, it's a right click/long press. This is a pretty painful process, I admit.

In the case of auto-discovery, there are two things to describe. First is how to get to the homepage for the blog. This is necessary because if the user lands on e.g. a Marginal Revolution blog post, the auto-discovery URL points at a feed for the comments, which is different from the feed icon on the page which points to a blog feed. There's a wrinkle that the homepage for the site is not always the root of the blog.

True - I was assuming that "go to the front page of the site" was simple but yeah, it's not always. However, I suspect a lot of blogs do have both the overall posts feed and the overall comments feed (if any) in the head of every page. I just checked the first four sites Google gave me for "most popular blogs" (Mashable, TechCrunch, Boing Boing and TPM) and it's true for all of them - you can use a post URL for auto discovery.

On the other hand, non-blog sites, like commercial news sites, are probably less likely to do this. (It wouldn't occur to me to subscribe to any of them via RSS but it takes all sorts!)

In the case of the Feed URL, you subscribe and either it works or it doesn't

In the case of auto-discovery, it will either work, or appear that it's worked but you've actually subscribed to the category feed or the post comments feed (so there need to be instructions to check for that case), or it doesn't work and you have to fall back to the Feed URL approach.

I've only used auto discovery (in recent memory) in Feedbin which tells you what the names of the feeds are that you can subscribe to, before you subscribe to them - which doesn't seem any more confusing than choosing the feed by using the names on a web page.

I'm not super concerned either way, it just seemed odd to me that you were treating something that was supposed to make this whole "just subscribe to a feed!" thing easier as an obscure "advanced" feature :)

philgyford commented 3 years ago

I'm not entirely sure where to put this but it feels a little relevant to this discussion. Google are updating FeedBurner, including "turning down" some features, including "Browser Friendly" which is their name for displaying a web-friendly version of feeds.

For posterity, here's their interface for configuring that:

Screenshot 2021-04-18 at 15 15 18

And here are the results:

Screenshot 2021-04-18 at 15 15 57

(An unusual entry from Pepys' Diary that contains a table, which explains the slightly odd formatting of the body.

dvikan commented 2 years ago

I have created a feed discovery service which kinda alleviates this problem:

https://discovery.thirdplace.no/?q=github.com/genmon/aboutfeeds

genmon / aboutfeeds

How to talk about feed auto-discovery? #13