Page MenuHomePhabricator

Late reply to early comment added in unexpected place in the discussion
Closed, DeclinedPublic

Description

If you join a larger discussion, and reply to an early comment, then your comment will be placed fairly high in the discussion, rather than at the end. This might sometimes be wanted, but it is usually not expected.

Reported by User:Gustave67 at https://fr.wikipedia.org/w/index.php?title=Discussion_Projet:Outils_de_discussion&diff=167640482&oldid=167640472 and also in early testing on the Beta Cluster.

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptFeb 24 2020, 1:26 AM
Pppery added a subscriber: Pppery.Feb 24 2020, 4:15 AM

Why is this "not expected"?

Jc86035 added a subscriber: Jc86035.EditedFeb 24 2020, 10:07 AM

@Whatamidoing-WMF, I think it would help if you were to provide an example of a page on which this happens, since I think this is already supposed to be handled for the case in which all of the existing comments are correctly parsed.

@Pppery, I think the type of situation being referred to is probably something like this:

Comment
: Reply 1
:: Reply 1-1
: Reply 2
:: Reply 2-1
: (Reply 4 gets posted here)

: Reply 3
:: Reply 3-1
: (Reply 4 should be posted here)

I would need an example to investigate this, it's already supposed to behave like you want.

That placement looks correct. Replies go to the end of the list directly beneath them.

The comment in the fr.wp example (https://fr.wikipedia.org/w/index.php?title=Discussion_Projet:Outils_de_discussion&diff=167640482&oldid=167640472) seems to ask for the opposite of what this task asks for. The reply was placed at the end, and Gustave67's expectation was that it would be placed "fairly high in the discussion".

Currently, our behavior is that if a comment already has some replies, then your reply will always be placed below the existing replies. Sometimes a user might want their reply to go above existing replies. But I don't think there's any way for us to guess what is wanted, and in my experience the usual convention is to reply at the end, like we do now.

Pppery removed a subscriber: Pppery.Feb 26 2020, 11:24 PM
ppelberg claimed this task.Feb 27 2020, 6:08 PM
ppelberg added a subscriber: ppelberg.
This comment was removed by ppelberg.

If you make three comments, each replying to the other, then you get this:

0
:1
::2
:::3

If you then reply to each of those (e.g., reply with "First" to comment "1"), in order, then you get this:

0
:1
::2
:::3
::::Last
:::Middle
::First

This isn't necessarily the order that you'd want to read this discussion in, but I don't see how the software could guess what I meant.

JTannerWMF added a subscriber: JTannerWMF.

This is waiting for an update from @ppelberg and should be ready for close out and revisited at another time.

As, @matmarex describes below, currently, the Reply tool will post comments as follows:

  1. Indentation depth: the Reply tool will post your comment at one level of indentation "deeper" than the comment you are replying to (e.g. if you are a replying to :Comment A, your comment, Comment E will be posted to the page as such: ::Comment E)
  2. Comment ordering: the Reply tool will post your comment (Comment E) beneath replies to (Comment A) that have been posted chronologically before yours (Comment B, Comment C, Comment D).

This seems to be in line with existing conventions and expectations and as such we consider this ticket to be proposing behavior that is counter to this. This conclusion/assumption could very well be wrong, but we will not know until more people are using this.

As such, we are going to move this to our backlog for the time being.

Currently, our behavior is that if a comment already has some replies, then your reply will always be placed below the existing replies. Sometimes a user might want their reply to go above existing replies. But I don't think there's any way for us to guess what is wanted, and in my experience the usual convention is to reply at the end, like we do now.

I don't think there is any other logically consistent ordering we could use without violating a bunch of other conventions, so I think we should decline this task.

ppelberg closed this task as Declined.Mar 18 2020, 3:43 PM

I don't think there is any other logically consistent ordering we could use without violating a bunch of other conventions, so I think we should decline this task.

Declining sounds good. To make sure this information [1][2] is accessible to people who are looking for it, I think we should create an FAQ/Help page for the Reply tool. Work related to creating that page can be found in this task: T247994


  1. T245960#5921908
  2. T245960#5960673