XMPP in_reply_to should be set to last notice that has been sent over XMPP
XMPP in_reply_to should be set to last notice that has been sent over XMPP
| Issue ID: | 1884 |
| Issue Category: | enhancement |
| Component: | api |
| Priority: | major |
| Status: | fixed |
| Assigned: | evan |
| Milestone: | 0.9 |
| Keywords: | context, xmpp |
XMPP in_reply_to auto-guessing should only consider notices you have seen. Right now StatusNet can reply to a totally bogus notice and causing wrong in_context.
Consider the following scenario:
@alice sends a dent to !identica. This dent is sent to all in the identica-group. Then she writes another dent with some personal stuff.
@bob sees her !identica dent, but not her personal one (because he is in the identica group), and neatly replies @alice.
StatusNet will in the following situation put @bob's notice as a reply to @alice's private message. It should rather reply to the last notice from @alice that @bob saw.

Updates
#1
In 0.9, XMPP messages will show the notice_id in the [] after the message, ex: "Bob says hi! [78]"
I also added a "reply" command that you can use like this:
"reply 78 why hello Bob"
How does that solution sound?
#2
Replying to [comment:3 candrews]:
> "reply 78 why hello Bob"
> How does that solution sound?
toooooo long.
how about
r ID ?
#3
"r" is a synonym for "reply" - you are welcome to use it.
#4
Replying to [comment:3 candrews]:
> In 0.9, XMPP messages will show the notice_id in the [] after the message, ex: "Bob says hi! [78]"
>
> I also added a "reply" command that you can use like this:
> "reply 78 why hello Bob"
>
> How does that solution sound?
It sounds great to me. With the synonym situation, it'll work quite well.
#5
As to the last point, I'm collecting material and will create a separate new issue for that. But the wildest I saw (today) was a dent (reading actual context not a reply at all) being attached to a nine-month-old dent (by the account it was addressed to - at least it got that part right).
Conclusion: adding a reply command has made correct linkage possible (but not easy), and it's not a real solution anyway as the (wild) guessing is still taking place.
#6
See also http://status.net/open-source/issues/2835
You can also subscribe to the
RSS feed for updates to this issue.