--> seth (~seth@home.metamumble.com) has joined #☠
<seth> alex: So we were mildly foiled in trying to get the inbox going
<seth> alex: We're going to wait on finishing the schema changes
<seth> And then I think the whole server codebase needs some refactoring
<seth> DefaultMethods.java being a perfect candidate... lots and lots and lots of duplicated code
<seth> Also would be nice to have a getMessages() that takes an arbitrary Criterion
<seth> Or Criteria
<seth> (so you can do different sorts of queries... we needed this for inbox)
<seth> Hibernate does have a Criteria query system as an alternative to HQL
<seth> Looks pretty reasonable
<seth> Though not sure how we'd serialize it over the wire...
<alex> ok
<alex> Will the inbox really be doing so many different queries though?
<alex> can't they ones that end up being used be hardcoded as operations?
<seth> alex: We can hardcode
<seth> alex: Mostly I don't want to end up w/ 20 cut-n-paste query operations in the base class
<seth> alex: Even if we don't expose this over XML-RPC
<seth> Whatever the reason, the server has turned into terrible spaghetti code
<alex> exposing the database to clients really ties us to one specific implementation
<seth> It just needs some refactoring
<seth> e.g.
<seth> We shouldn't be doing the
<seth> try {
<seth> HibernateUtil.getSession
<seth> etc
<seth> stuff
<seth> In every call
<seth> We should implement our own introspection handler and do that and then pass in the session
<alex> that would be nice, yes
<seth> alex: yeah
<seth> alex: I'm more concerned about the internals than exposing it to clients
<alex> public Object execute(String methodName, Vector params) throws Exception {
<alex>  
<seth> We just haven't been cleaning up as we go along
<alex> Can't that just do the hibernat session thing?
<seth> alex: yup
<alex> I'll do that then
<seth> alex: But we also need our own introspection thing to handle return values
<seth> alex: We shouldn't be doing type conversions in each method
<seth> alex: I briefly looked to see if the XmlRPC library had a way of registering these
<seth> alex: Didn't see one off-hand
<seth> I think we just have a general problem in the server that we've been implementing functions by cut-n-paste rather than figuring out how to share code between them
<seth> Obviously we don't want to take this too far... but there's a lot of unneeded duplication right now
* seth shrugs
<seth> The client is way more messy atm though
<alex> well, we've been focusing on experimenting and features, so design and cleaness suffer
<alex> no big suprise
<seth> yup
<seth> Well, also its more people working on it now
<alex> nor any big harm
<seth> Hence less likely to recognize when cleanups are needed
<seth> yup yup
<seth> In the spirit of experimentation though
<seth> It does suck that DefaultMethods are all very long
<seth> And all very similar
<seth> And in varying degrees of use/bitrot
<seth> So inbox wise
<seth> We realized we need to know *when* people subscribed
<seth> To a particular topic
<seth> Or forum or whatever its called now
<alex> ok
<seth> (they're not actually forums, fwiw, they're supposed to be much much more lightweight)
<seth> (I know its an implementation detail, but the name worries me)
<alex> (subscription data is not stored in the server atm)
<seth> yeah
<seth> We probably have to store it on the server after your change
<alex> yeah
<seth> s/have to/should/
<alex> so, what name would you prefer over forum?
<seth> But clearly, you don't want a bunch of new messages appearing yesterday when you subscribe to something today
<seth> alex: topic
<alex> and a topic?
<alex> thread?
<seth> eh?
<seth> There's only one topic.
<seth> Anything else is blasphemy.
<alex> whatever
<seth> So we don't have heirarchy, right
<seth> We just have links where a topic was broken out / created
<alex> What name would you use for the object that in the new_design branch is called Topic?
<seth> Uh... what does it do?
* seth didn't notice that
<alex> http://people.redhat.com/alexl/yarrr/new_proposal.txt
<seth> oh
<seth> No forums
<seth> Just topics
<seth> Or no topics
<seth> Just forums
<seth> The current client is just a random ass thing somebody put together
<seth> Its not really how we should be doing things
<seth> This stuff has all sort of been waiting on inbox
<seth> We need links embedded in arbitrary points in topics
<seth> Interleaved w/ messages
--> seth_ (~seth@home.metamumble.com) has joined #☠
<seth_> alex: server split
<seth_> If you said anything I didn't get it
<alex> i didn't
<seth_> I just sent you message w/ all my typing that got lost
<alex> message?
<alex>  /msg?
<seth_> email
<alex> ah
<seth_> Basically: the current client UI makes a very arbitrary distinction between different kinds of topics that I don't want (toplevel and non-toplevel)
<seth_> And Forum vs. Topic takes that even further
<-- dhm_away has quit (irc.acc.umu.se ircd.gimp.org)
<-- seth has quit (irc.acc.umu.se ircd.gimp.org)
<-- jrb_sleep has quit (irc.acc.umu.se ircd.gimp.org)
<-- gicmo has quit (irc.acc.umu.se ircd.gimp.org)
<seth_> The sky is falling, the sky is falling
--- seth_ is now known as seth
<alex> heh
<seth> Oh the two of them, I think Topic is "more right" than Forum
<seth> s/Oh/Of/
<seth> As long as we have the ability to interleave links to other topics with messages inside the topics
<seth> I think just having Topic will be fine
<alex> So, you find a bug in nautilus that you want to ask about
<alex> what do you do?
<seth> I don't really want to embed the idea of a "core group" in the software
<alex> create a topic and post in it?
<seth> You go to the nautilus website
<seth> And it has a link to the nautilus topic
<seth> And you post a question
<seth> And three people respond
<seth> And somebody goes and makes a topic out of it
<seth> Or you create a new topic
<seth> And its linked into the nautilus topic
<alex> so, we're into the hierarchy again?
<seth> I never had problems w/ the heirarchy ;-)
<seth> Though its not strictly heirarchical
<seth> Its a graph
<alex> nobody ever could describe how it would look in a client
<seth> eh?
<seth> You're thinking browsing
<seth> Inbox is easier
<seth> You remember...
<seth> New stuff comes in
<seth> Notices of a new topic being linked into a topic you read (say nautilus)
<seth> You click to get that topic too...
<seth> I mean, browsing is important
<seth> But I'm a lot less concerned about it
--> dhm_away (~dmalcolm@nat-pool-bos.redhat.com) has joined #☠
--- irc.gimp.ca gives channel operator status to dhm_away
<seth> For all I care, the main access to browsing could be search
<seth> And then links work just like the web
<seth> No real "heirarchy" per se
<alex> i don't think searching can replace browsing really
<alex> they are sort of different operations
<alex> "find a solution to problem" vs "see what these people are doing these days"
<seth> Nobody has defined why you need browsing at all
<seth> When I say browsing I mean
<seth> "Accessing content you are not watching"
<seth> The most obvious way is it becomes relevant, say on IRC, and somebody pastes a link
<seth> In which case the problem is moot
<alex> not really
<seth> The other way is that I want to find out something about... something... in which case search makes some sense
<alex> someone posts about a new project
<alex> that i get interested in
<alex> i often browse their lists to see what sort of people they are, and what sort of stuff they are doing
<seth> right
<seth> So...
<seth> You search for the project
<seth> Come up w/ its topic
<seth> And look through that
<seth> And it'll have links off to any topics (many presumably) that branched off that
--> jrb_sleep (~jrb@cpe-66-189-91-28.ma.charter.com) has joined #☠
--> gicmo (~gicmo@www.deep-sorrow.de) has joined #☠
--- ircd.gimp.org gives channel operator status to jrb_sleep gicmo
<alex> so, "linked off topic" somehow means a thread that got big?
<seth> I've always thought of this form of yarrr as being more like a web browser than ... something formal
<seth> sure
<seth> Or some people will turn every thread into a topic I suppose
<seth> Just depends
<seth> Like I can imagine every thread on desktop devel list except the occasional announcement being a topic
<alex> Thats how the Forum/Topic thing came up
<seth> s/thread/discussion/
<seth> I don't see a need for a distinction
<seth> There's a difference at the extremes
<seth> But I want fluid mixing in the middle
<seth> FWIW, summary shouldn't really be "per topic"
<seth> Summaries are "for these messages" (could be a single message)
<seth> Summary != Description... exactly.
<alex> I think the difference came up with our browsing based client. We weren't able to come up with any sane way to display something that had both "subtopics" and message
<alex> so, summaries wouln't be of a whole thread, just some subset of it?
<seth> yeah
<seth> UI wise
<seth> Well, using Evolution style display
<seth> Select messages (shift-select, or ctrl select a couple)
<alex> and someone else could summarize a different subset (instead of say extending the current thread summary)
<seth> Right click
<seth> "Summarize"
<seth> Sure, possibly
<seth> Though I might group messages that were summarized in the sort order
<seth> We'd need to play with that
<seth> The current "group by threading"
<seth> I'm not so confident in
<seth> There are advantages to a linear list and using different groupings
<alex> i dunno about that
<alex> it makes summarizing some messages so similar to just replying to it
<seth> Except that summaries "stand in" for the message
<seth> Its an alternative to picking exemplar messages in a thread
<seth> It allows people to write exemplar messagse
<seth> That are supposed to be "NPOV"
* alex isn't at all sure that will work (people being neutral that is)
* seth not sure it will either
<seth> That's why we need people using this
<seth> So we find out what works and what doesn't
<seth> But unfortunately email feeds is taking a while to get done
<seth> Summaries being neutral enough to be useful will probably require a little culture work
<alex> well, email feeds are of a different type of structure than what you're talking about for yarrr. Importing them sort of forces the email structure on yarrr.
<seth> If we shape the tool right it might happen naturally...
<seth> I hope not!
<seth> I want to force the email to fit in yarrr
<seth> Tossing email-y things when they don't
<alex> well, since the yarrr structure isn't well defined yet, working with email import is likely to make the structure we develop email:ish
<alex> So, I'm not sure what to do today
<seth> I'm not sure what things you're referring to?
<seth> uhm
<alex> I was gonna finish my work, which is lots of commenting, testing etc
<seth> Well, what might you do?
<alex> like, fixing all the tests to work in the new structure.
<seth> That sounds reasonable, sans Forum
<seth> yeah
<seth> We're waiting on the inbox until your changes are done
<seth> That's probably the big UI change we need to start situating a bunch of features we've been holding off on
<seth> Since the browsing interface shouldn't really be the primary
<alex> well, you say "sans forum" as if that doesn't make the whole new design irrelevant
<alex> or at least needing to be changed a lot
<seth> ah
<seth> I don't really know what the new design requires?
<seth> Why isn't just Topics with linking ok?
<seth> The client needs UI change
<seth> alex: Still there?
<alex> the linking part isn't actually done yet
<alex> there are just messages in topics
* seth mildly wanting to go to bed but will talk as long as he has to
<seth> So maybe the linking is part of what we need
<seth> Its *all* links anyway
<seth> Since messages aren't "in" topics
<seth> In some ways
<seth> Or rather... heirarchy, links, what is the difference?
<alex> the difference is all in how you display it of course
<alex> links tend to replace what you followed, while hierarchies tend to be displayed inline
<seth> yup
<alex> CS wise they are of course both just arrows between nodes
<seth> So...
<seth> Back to "what to do"
<seth> um
<seth> I'm not clear on what's needed
<alex> Neither am i
<seth> Or rather, how can I help you figure out what you need to do?
* alex is totally confused
<seth> In truth, I didn't fully understand the need for the redo of the schema
<seth> Probably partly because the existing model was very roughly what I thought made sense for what I was picturing
<seth> Though we seem to have been on different pages
<alex> partly it was to add support for user data
<seth> But I also haven't been looking that closely because of CS201 messes, and cube reorg
<seth> Yeah
<seth> I think that is awesome
<seth> (I'm not saying I don't think there wasn't a reason, sorry if that wasn't clear, just that I never sat down and figured out what it was)
<alex> and a sane way for the client to get incremental updates
<seth> I think that's focusing too much on minutia, though probably good overall
<seth> I can live with a few race conditions right now ;-)
<alex> the incremental updates part made the whole DAG-of-topics really complicated
<seth> I'd rather have the DAG-of-topics than really clean incremental updates
<alex> I think much of it is because we still think of a browser-like ui
<seth> yeah
<alex> because i don't think anybody has really grasped what it would look like otherwise. I mean, we have some ideas and notions, but not really a coherent picture of it.
<alex> I can remove the forum part and try to come up with some good way of doing links instead
<seth> Coherent picture comes from trying
<alex> it'll sort of break the current client ui, but we can redo that later
<seth> We don't throw away nearly enough code
<seth> Not that that's a goal
<seth> But we're still engineering thinking the code we write is going to be used in some shape or form
<seth> Havoc had a good idea today, mostly joking
<seth> Which is we write everything in Flash
<seth> So we *know* we're not going to use that code and we can just try shit
<alex> see, the problem is that at the same time we're saying we're gonna use it, and we need to keep our data around...
<alex> i understand that we need to use it to know what works and what not
<seth> yeah.
<alex> but keeping the data around makes it hard to experiment with the structure of the data
<seth> That is a problem.
<seth> One solution is to be less attached to the data
<seth> Honestly.
<seth> dhm is *obsessed* with it
<alex> i'm all for that
<alex> i don't understand it really
<seth> He refuses to post anything unless it'll be archived for eternity
<seth> And I'm all for using email for really important messages still
<alex> i think it was a mistake to not have a mailing list initially
<seth> But for normal project chatter, I don't really care if it dies every week or so
<seth> yeah
<seth> you are totally right
<seth> my bad
<seth> I mis-estimated how long it would take to get people pulling forward on stuff
<alex> nobody knows what the heck people are doing
<seth> And get yarrr pulled to where it was nice to use
<seth> So part of the problem is that jrb disappeared forever at exactly the wrong time
<alex> yeah, that was bad
<seth> And even now he's not fully back in the saddle
<seth> Though I think he came a lot closer
<seth> I've been pair programming with him when I have time
<seth> And we keep randomly losing people to errata and stuff
<seth> Though hp claims that should be over with now
<seth> timezone is definitely a killer too :-(
<seth> I notice a big improvement whenever we overlap on IRC for a couple hours
<seth> I dunno.
<alex> everyone is working at sort of fringe issues, mail import, database updates, etc. But those are hard to do when there is no common work and decisions on the core
<seth> What is the core?
<alex> and there is no such work because none of our discussions (like this one) is stored so others read it
<seth> I disseminate
<seth> In person
<seth> When we have these discussions
<alex> the core is "whats the idea of how users would use yarrr"
<seth> You<->Marco   and Seth<->jrb<->dhm<->walters are sort of two islands right now
<seth> ah.
<seth> The answer is
<seth> "That's what we're trying to figure out"
<seth> This isn't a design like when you can just copy somebody!
<alex> sure
<seth> With that its easy to draw everything out from the start
<seth> But to do this sort of thing
<seth> We need better communication
<seth> So we don't ignore the core issues
<seth> Is that what you're saying?
<alex> but i find we sort of loose the idea where we were, because you can't like read the old discussions, or be sure that everyone read the same discussion
<alex> seth: something like that
<seth> I'm happy to have a mailing list
<seth> If we need one
<seth> I guess we might
<alex> Then we could import it into yarrr :)
<seth> And the day when we prefer yarrr to the mailing list is the first step toward victory :)
<alex> exactly
<alex> so, about links
<seth> I do think we don't really have a problem today
<seth> If we were more concerned w/ having the thing running
<alex> do you seem them as sort of a messsage?
<seth> And less concerned with storing data
<seth> We could store "sent messages"
<seth> Easily
<seth> so you could repost anything quickly that might have been lost
<seth> The inbox will remov ethe "what has actually changed" problem, though your user data thing could fix that too
<seth> alex: erm...
<seth> alex: sure?
<seth> alex: Well, its all just stuff in a topic
<seth> alex: I'd want a summary of the topic
<seth> alex: How many messages
<seth> alex: Maybe some of the names of people who'd posted there
<seth> alex: How many messages in the topic I'd read vs not read
<seth> All in one line
<seth> Like a collapsed message
<seth> alex: so um
<alex> but it'd be "mixed in" with normal messages?
<seth> alex: I don't want to leave you dangling
<seth> alex: But its 5am
<seth> alex: yeah, mixed in
<seth> Maybe sorted to wherever the "last date" of a contained message was
<seth> alex: Any last things I could do to help you have something useful for today?
<alex> i don't think so
<alex> i'll try to think about this for a while
<seth> ok
<seth> excuse the mess ;-)
<seth> alex: Please email me if you have ideas for stuff I can do to more concretely help
<seth> maybe we need some step through experiences
<seth> They can clash and stuff
<seth> But just some ideas about user experiences we might want to support
<seth> I feel like we've already talked a lot about this
<seth> But
<seth> It never seems to stick
<seth> so...
* seth shrugs
<-- seth has quit (Leaving)