1261540 - Reader mode not offered on some pages (e.g. WHATWG specs) because of 'home page' URL filtering on the Firefox side

Status: NEW

(Toolkit :: Reader Mode, defect, P5)

Description

[Domenic Denicola]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:45.0) Gecko/20100101 Firefox/45.0
Build ID: 20160315153207

Steps to reproduce:

Browser to https://streams.spec.whatwg.org/ in one tab, and https://streams.spec.whatwg.org/index.html in another


Actual results:

In the first tab, reader mode is not offered. In the second, it is.

This is because of the "home page" check in https://mxr.mozilla.org/mozilla-release/source/toolkit/components/reader/ReaderMode.jsm#334


Expected results:

Reader mode should have been offered in both.

This effects all other specs listed at https://spec.whatwg.org/ (e.g. HTML, DOM, URL, etc.)

Comment 1

[:Gijs (he/him)]

(In reply to Domenic Denicola from comment #0)
> This is because of the "home page" check in
> https://mxr.mozilla.org/mozilla-release/source/toolkit/components/reader/
> ReaderMode.jsm#334

Yes, unfortunately getting rid of this would break lots of more commonly used sites for the general web population (that is, we'd offer reader mode on homepages of nytimes and the like when we shouldn't). I don't really have any amazing ideas about how to fix this. There are a number of constraints, including not regressing loading performance (we can't really spend more than 2-3ms evaluating whether a document should offer reader mode when loading it). If you have suggestions, they'd be very welcome! In the meantime, going to downprio this because we should fix even grosser offences first (like dropping entire paragraphs/sections out of CNN articles etc.).

Comment 2

[Domenic Denicola]

Maybe you could consider searching for a specific meta tag, e.g. <meta name="offer-reader-mode" content="true">? (I guess content="false" could be used as an opt-out for incorrect detections... although maybe that would just be abused by pages that don't want reader mode hiding their ads.)

Comment 3

[Smylers]

This also seems to be affecting Hixie's blog (though I can't find the default page name to verify it with an alternative URL). All individual articles are reached with URL parameters on ‘the homepage’:
http://ln.hixie.ch/?start=1700627373&count=1

Maybe the homepage detector could only trigger when the path is / and there is no query part? Or where the query part only contains utm_foo tracking parameters?