#social 2019-06-02

2019-06-02 UTC
# 02:49 
trwnh fr33domlover: it should include direct descendants only
# 04:48 
fr33domlover trwnh, explain? :)
# 04:49 
trwnh fr33domlover: post a has replies b, c, d , e. post b has replies f, g, h. and so on.
# 05:02 
fr33domlover trwnh, well yeah that makes sense, but the other approaches make sense to me too... I'm looking for a way to *decide* ^_^
# 05:02 
fr33domlover Even if to decide simply means to choose a convention
# 05:04 
trwnh fr33domlover: other ways seem wrong...
# 05:07 
trwnh fr33domlover: let's put it this way. the "replies" collection is supposed to keep track of objects which have inReplyTo targeting that object
# 05:19 
fr33domlover trwnh, well that's why I'm asking: should we treat replies as the reverse property of inReplyTo? In many systems, e.g. on Discourse, replies are displayed as a list, with link to parent when inReplyTo is another message in the thread. I don't mind deciding to do things the way you're saying, I just want to get more people's thoughts so that we can agree on this together
# 05:22 
trwnh i dont see why it should be any other way
# 05:23 
trwnh replies: Identifies a Collection containing objects considered to be responses to this object.
# 05:23 
trwnh Range: Collection
# 05:25 
trwnh that's directly from the activity vocab spec
# 05:27 
trwnh replies: {type: Collection, items: [{},{},...,{}]}
# 05:28 
trwnh or items: [@id, @id, ..., @id] -- you get the point
# 05:31 
fr33domlover trwnh, have you looked at Discourse, or GitLab, or githu8, etc.? When they display comnents, they display them all as a list, and they just link between children and parents. In that sense, all descendants are replies on the topic/issue/PR, and inReplyTo is secondary detail. I'm not saying we should prefer that approach for AS2 'replies'; just explaining how it can make sense too
# 05:31 
rialtate[m] fr33domlover: context is generally the flattened list
# 05:33 
trwnh the display is irrelevant from the structure
# 05:33 
trwnh you can simply crawl all children and expand the tree then render it however you want
# 05:34 
fr33domlover rialtate, I'm already using context for something else ^_^ but thanks for the feedback! I'll ask around a bit more to be sure everyone-ish agrees on the direct children approach
# 05:34 
fr33domlover trwnh, true
# 05:34 
trwnh what are you using context for?
# 05:34 
trwnh i would imagine context as the issue itself
# 05:35 
fr33domlover trwnh, yeah, to link from all comments to the issue/PR/etc.
# 05:35 
trwnh and then you can reply to the OP for everything after the issue description
# 05:38 
trwnh honestly context is unnecessary even, you can just treat the root post as the issue itself
# 05:40 
rialtate[m] I'm not sure that it matters which convention you use for replies since it is merely used as a discovery tool and @id's deduplicate while inReplyTo gives you your linkages
# 06:00 
fr33domlover rialtate, yup, it doesn't matter much, I just want to officially pick a convention :) I'll add it to my list of proposals to make it explicit I'm choosing and proposing it
xmpp-social joined the channel
# 07:50 
trwnh im still not sure what the other proposal even is
# 07:51 
trwnh i mean, sure, you can claim whatever you want in the "replies" Collection... but imo it's not valid to list an Object as part of "replies" if it's not also "inReplyTo"
# 07:51 
trwnh i would personally treat that as invalid
# 07:52 
trwnh if you embed transient objects in there then whatever
# 07:53 
trwnh but if you have objects with an id then it would make sense to fetch those objects and then check the inReplyTo to verify that they are actually replies
# 07:56 
trwnh also if you say that a has direct replies b,c and b has direct replies d,e,f and then you say a's replies are bcdef then you're setting yourself up for exponential Updates
# 07:57 
trwnh because now replying to any object will have to Update not just that object, but also every object in the chain above it
timbl joined the channel
# 08:24 
rialtate[m] trwnh: Updates isn't really a fair criticism since they are inherently problematic no matter what
# 08:26 
trwnh sure it's fair. it actually doesnt matter that update is expensive (although that does make it worse)
# 08:26 
trwnh the general point is that one is a constant-time operation
# 08:26 
trwnh and the other is not
# 08:27 
trwnh before even doing any delivery
# 08:28 
trwnh 1 operation vs who knows how many (depending on how deep you are in the tree)
# 09:09 
fr33domlover trwnh, I'm not sure I see how updates would become expensive
# 09:09 
fr33domlover It just means that if other servers cache replies, they should rely only on caching the direct children
# 09:09 
fr33domlover And not the rest
# 09:10 
fr33domlover And given that, no need for exponential updates
# 09:10 
fr33domlover Either way, I'm probably going for the only-direct-replies approach
Enot and Enot_ joined the channel
# 15:22 
Enot_ Has anyone noticed the test suite site having clock related certificate issues?
# 15:24 
fr33domlover Enot_, yeah, it's been like this for a while
# 15:24 
Enot_ That's a shame. I'd love to use it. Is there another way to run the test suite?
Robbie joined the channel; Enot_ left the channel
# 15:34 
enot Or do you know who manages the server for the test suite site?
timbl and Guest84 joined the channel
# 16:54 
cjslep[m] I think Chris is well aware, but busy. :(
timbl, rZr and Clipsey joined the channel