On 16 Aug 2010, at 16:00, Dave Cridland wrote: > On Sun Aug 15 12:21:15 2010, Candide Kemmler wrote: >> I'm very interested in the PEP xep, in particular in combination with the >> User Activity XEP. However, I have a few remarks about the specification as >> it is. Also, maybe it is just me, but I can't seem to find any meaningful >> implementation of PEP among existing clients and servers. The best I could >> do is with tigase and psi: I enabled the "Publish tunes" in the psi client >> and I could indeed see that the xml was pushed to another client in the XML >> console. But that's about it: an XML message in a debugging console. I think >> pep is pretty cool and I wonder why it's not given a bit more attention from >> implementors. > I think it depends on the client. I personally find Gajim is doing very well > here, publishing and displaying information as the "rich presence" it was > intended to be. >
Unfortunately, it turned out to be a real pain to try to run Gajim on my Mac, so I will have to trust your word for it :( > >> Now to my remarks: >> First off, it is my feeling that providing a fixed activity list, even if it >> can be supplemented by way of a catch-all <other/> category, is very >> arbitrary and that many people might feel very reluctant to log their >> activities as dictated by a such an authoritative list. Further organizing >> that list in a two-level hierarchy is also very constraining: why wouldn't >> people want to organize their activities in much deeper hierarchies? I had a >> look at onesocialweb's attempt to use activitystreams for that, but that >> format seems to be bound to the same limitations. > I think "machine readable" is the constraint here. If you've an alternate > form that manages to provide machine processing and yet allows greater > expressivity, this would be interesting. Since I gave up, I no longer have > the major problem of not being able to say I'm drinking beer *and* having a > smoke, though, so I'm not so concerned... > IMO User Activities can be interesting for instrospective behavioral analysis. I mentioned a couple use-cases: substance abuse, eating disorders or more commonly, exercising... There are user activities that are potentially fascinating to track. But that would probably need some marketing, i.e. I don't think that current generic jabber clients are going to make a very good job at explaining how to best take advantage of that feature. Nonetheless I find that the decentralized nature of xmpp makes it a perfect fit for privacy aware users. > >> Now to something more important: as I understand it publishing activities is >> an all-or-nothing option: subscribers can optionally filter out activities >> they are not interested in, but I cannot myself choose which activities I'd >> like to share with a specific friend or group of friends. That's too bad: I >> was thinking of activities as a way to share statistics - sports statistics >> for example, and I don't want to potentially bother everyone with these >> statistics (or force them to filter them out). Another use would be for me >> to share some bad habit, like alcool, smoking or drug use for example, while >> seeking support from a (potentially anonymous) group of people who share the >> habit. In that case, I obviously wouldn't even want others to know that I >> have that kind of behaviour. > Well, the activity node *can* be filtered out to only specific people, but > it's not often a feature available in the server - changing the access model > and configuration of the PEP node will allow this. > As I understand it, pubsub allows several access models for nodes, one of them being for example "whitelist". Yes, but that still means that everyone can _see_ the node, hence anyone would see that I'm tracking, let's say, my eating habits, and I don't want that. > It won't, however, allow anonymous activity information - for that, you > really need to send the events via some anonymizing system, like a MUC. > > I'd also note that specialized information really means a specialized client > (or rather, one that understands the PEP node you're trying to use), and PEP > makes that easy to do. > That is very clear. > >> My third remark is on statistics: aside from the type of activity, there's a >> lot of additional info that users are likely to be wanting to track, if they >> bother to log their activity in the first place: let's mention: duration of >> an exercising session, amount spent on a meal, title of book being read, >> etc... It would be great to allow people to add such information. > I don't think there's anything preventing you from adding extra information > there, except that you may well not want to receive all that information. > > I do quite like the notion of advertising the book I'm reading, for instance, > but I wouldn't put that in User Activity, I'd put that somewhere else - much > like User Tune isn't bundled into Activity. Well then, that opens up for a whole infinite list of XEPs, one for every kind of activity. That's essentially the problem I'm trying to solve: IMO "Personal Eventing Protocol" should allow people to state all sorts of events that are affecting their lives, according to _them_. The key word here is probably semantics. There must be a way for users to create their own semantics for events they deem enough important for them to track. I'm thinking along the lines of _basic_ semantic web stuff, like: users should be able to create statement types, like say: "reading" [book_title] [book_author]? [isbn]?, or "ran" [distance] "miles", etc... then share the statement types so other user can use them as well if they want to. > Dave. > -- > Dave Cridland - mailto:d...@cridland.net - xmpp:d...@dave.cridland.net > - acap://acap.dave.cridland.net/byowner/user/dwd/bookmarks/ > - http://dave.cridland.net/ > Infotrope Polymer - ACAP, IMAP, ESMTP, and Lemonade