1209683 - Reader Mode and RM first time use bubble should not be available on Pinterest

Status: RESOLVED DUPLICATE of bug 1210366

(Firefox :: General, defect)

Description

[Verdi [:verdi] Best to slack me]

STR:
1. Create a new profile
2. Visit Pinterest
3. Click on reader mode
4. Reader mode fails with error, "Failed to load article from page"

Comment 1

[:Gijs (he/him)]

I'm fixing the tour aspect with the blocklist that bug 1210366 decided on.

For the reader mode offer... The problem here seems to be that there are loads of .pinDescription paragraphs with actual text (like the descriptions of pinned items). There's no way for the isProbablyReaderable code to see that they don't form part of an overarching story without making that code a lot more slow, which we don't really want to do...

Margaret, do you have any clever ideas?

Comment 2

[:Margaret Leibovic]

(In reply to :Gijs Kruitbosch from comment #1)
> I'm fixing the tour aspect with the blocklist that bug 1210366 decided on.
> 
> For the reader mode offer... The problem here seems to be that there are
> loads of .pinDescription paragraphs with actual text (like the descriptions
> of pinned items). There's no way for the isProbablyReaderable code to see
> that they don't form part of an overarching story without making that code a
> lot more slow, which we don't really want to do...
> 
> Margaret, do you have any clever ideas?

Like you said, I'm not sure I know of a great solution here... maybe we should somehow inspect the ratio of images to text? Pinterest is essentially a lot of images with captions, but the caption text is what we're detecting as "readable" content.

I was tempted to say something like looking for adjacent paragraphs to text to signal an article, but like you say, that would make this a lot slower, and authors can create arbitrary nested mark-up. Maybe the depth of the paragraph elements themselves could indicate whether or not it's an articles? But they could be pretty deep on complicated news sites.

Comment 3

[:Gijs (he/him)]

(In reply to :Margaret Leibovic from comment #2)
> (In reply to :Gijs Kruitbosch from comment #1)
> > I'm fixing the tour aspect with the blocklist that bug 1210366 decided on.
> > 
> > For the reader mode offer... The problem here seems to be that there are
> > loads of .pinDescription paragraphs with actual text (like the descriptions
> > of pinned items). There's no way for the isProbablyReaderable code to see
> > that they don't form part of an overarching story without making that code a
> > lot more slow, which we don't really want to do...
> > 
> > Margaret, do you have any clever ideas?
> 
> Like you said, I'm not sure I know of a great solution here... maybe we
> should somehow inspect the ratio of images to text? Pinterest is essentially
> a lot of images with captions, but the caption text is what we're detecting
> as "readable" content.
> 
> I was tempted to say something like looking for adjacent paragraphs to text
> to signal an article, but like you say, that would make this a lot slower,
> and authors can create arbitrary nested mark-up. Maybe the depth of the
> paragraph elements themselves could indicate whether or not it's an
> articles? But they could be pretty deep on complicated news sites.

Yeah. At this point, we updated bug 1210366 to just block reader mode for the site. I don't think that I have time to rearchitect things here, and it seems like it would be very tricky to do... so I'm just going to dupe this to that bug. :-(