AW Groupies/Chat Logs/AWGroupies-2010-07-27

[09:30] Mojito Sorbet: wxWidgets is being obstinate. A program in the demo package works fine, but when I try to compile my own copy of the same code, it fails at runtime
[09:34] Rex Cronon: hello everybody
[09:34] Honour McMillan: Good morning :)
[09:35] Rex Cronon: hi honour
[09:35] Rex Cronon: we should call this meditation circle:)
[09:35] Honour McMillan: :)
[09:36] Zha Ewry: /
[09:36] Zha Ewry: Hello all
[09:36] Rex Cronon: hi zha
[09:36] Honour McMillan: Hello :)
[09:37] Rex Cronon: u came to join the meditation circle, zha:)
[09:37] Zha Ewry: Morgaine has proposed a topic for today, which came out of the brief discussion on VWRAP last week
[09:38] Zha Ewry: /me suspects from silence Morgaine is off argying with the tea kettle
[09:39] Morgaine Dinova: Back
[09:40] Morgaine Dinova: Sorry about that.
[09:40] Zha Ewry: Hows the tea?
[09:40] Morgaine Dinova: Actually was drinking some perl
[09:40] Morgaine Dinova: Hmmmmm .... what was it I was proposing?
[09:40] Zha Ewry: Don't drink that hot, it will scar your throat like Sake
[09:40] Zha Ewry: The events stuffs
[09:41] Morgaine Dinova: Oh yeah, scalability assist for each cap.
[09:41] Mojito Sorbet: Like a 402?
[09:41] Morgaine Dinova: Oooh, I love hot sake. Not had any since Xmas
[09:41] Mojito Sorbet: oops, I think I mean 300-simething
[09:41] Zha Ewry: Lovely way to get throat cancer, alas, but tasty
[09:41] Zha Ewry: So, Morgaine in response to Meadhbh's latest
[09:42] Zha Ewry: was pushign a little on the question of how realtime caps get in
[09:42] Mojito Sorbet: 303 is the one I was thinking of
[09:42] Morgaine Dinova: Bah, spoil sport Zha, hot sake is worth mere cancer :P
[09:42] Zha Ewry: My take for quite a while, has been (and it shows in the client side caps stuff)
[09:42] Zha Ewry: that when you ask for a cap
[09:43] Morgaine Dinova: Well first, in general about mead's proposal, I think that's great. Sounds like she's able to move more freely outside of LL
[09:43] Zha Ewry: There is a good chance the service may ask to deliver the results in another channel.
[09:43] Dzonatas Sol: Well, maybe...
[09:43] Zha Ewry: Killing LLIDL, effectivlye, and focusing on the LLSD level description?
[09:43] Dzonatas Sol: if a service tries to use a cap, and the cap doesn't exist, then it may result in another channel
[09:43] Mojito Sorbet: I dont hink this has anything to do with the low level coding
[09:43] Rex Cronon: zha. r u listening to group chat?
[09:44] Zha Ewry: Not closely
[09:44] Morgaine Dinova: Yes Zha, it won't help when the service is external and the cap has no special knowledge of the service. But as an option for when they're co-located, it's pretty uber.
[09:44] Zha Ewry: Well, so the overall pattern is
[09:44] Zha Ewry: I want to get service "string which is opaue"
[09:44] Mojito Sorbet: This is a high level thing. It sounds to me not unlike the 300-series of HTTP status codes, where a server can redriect a requesting client to a different URI
[09:44] Zha Ewry: and the cap granter passes you back something
[09:45] Zha Ewry: There is NO requirement it be in the same DNS or even the same transport binding as the cap request
[09:45] Zha Ewry: so, I can ask for "Local.chat.services"
[09:45] Morgaine Dinova: I'm happy with opaque string, since the raw low level IP:port is a subset of a URI
[09:45] Zha Ewry: and get back a RTP or a XMPP thing to play with
[09:45] Zha Ewry: Well, this is the inbound side "I want to do X"
[09:45] Zha Ewry: on the response
[09:45] Mojito Sorbet: and get back the NAME of such a thing.
[09:45] Zha Ewry: the service granter gets to say
[09:46] Zha Ewry: Here is how I do it
[09:46] Zha Ewry: and.. then.. (and this is why client side caps) and why events / realtiem gets interesting
[09:46] Zha Ewry: There is a need to say
[09:46] Zha Ewry: I'd like to delver you stuff on pipe "arbitrary"
[09:47] Morgaine Dinova: Yep, very much so
[09:47] Morgaine Dinova: Hiya Dahlia :-)
[09:47] Zha Ewry: where... that is, in my mind, a URI and there is some fun negotation bits to make sure that the server and client pick a workable set for them, ideally best available
[09:47] Dahlia Trimble: hi :)
[09:47] Zha Ewry: The two ends are losely coupled, by definition
[09:47] Rex Cronon: hi dahlia
[09:48] Mojito Sorbet: Loook at this: http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html
[09:48] Mojito Sorbet: And scroll down to section 300
[09:48] Zha Ewry: so.. you may have to negotiate al ittle to make sure you have a pair in common
[09:48] Mojito Sorbet: Now this is written in webbie speak, but the concept I think is generlaly applicable
[09:48] Mojito Sorbet: In particuolar see code 300, the "multiple" response
[09:49] Zha Ewry: Yes and no
[09:49] Zha Ewry: Because it's not a 300 type redirect to another URI exactly
[09:49] Dzonatas Sol: Zha, the way the transport mehods are being compared to a device driver with various hardware addresses possible i thought was an insightful way to look at it
[09:49] Mojito Sorbet: Well, ONE of the 3xx codes seems to be like this
[09:49] Dzonatas Sol: except maybe everybody nogt familiar with device drivers
[09:49] Morgaine Dinova: What concerns me most is that it be possible to have the leanest, meanest form of transport when desired. I don't mind if it takes a while to negociate that transport via a zillion layers (gotta shore up Intel's CPU-selling business after all), but what I DON'T want is that the high overheads are present on the transport pipe once it's negaociated.
[09:50] Zha Ewry: and totally agree Morgaine
[09:50] Zha Ewry: We can do an expensive setup step
[09:50] Zha Ewry: That's fine
[09:50] Dahlia Trimble: nice landing Latig ;)
[09:50] Latif Khalifa: i was practicing ;)
[09:50] Zha Ewry: Once our two endpoint have decided to use hypersonic carrier pigeions
[09:50] Dahlia Trimble: *Latif
[09:50] Morgaine Dinova: lol
[09:50] Zha Ewry: we just let them get on with it
[09:51] Zha Ewry: Its not quite a 300
[09:51] Morgaine Dinova: Should install a trampoline for heavy landings, bounce them right back out to next-door sim :P
[09:51] Zha Ewry: since a 300, rougtly speaking is "This RESOURCE is now at URI X"
[09:51] Mojito Sorbet: WHich, for this session, it is.
[09:51] Zha Ewry: (Depending a little on what you mean by resource and URI)
[09:51] Zha Ewry: For this client/resource pair it is
[09:52] Zha Ewry: 300
[09:52] Mojito Sorbet: yes
[09:52] Zha Ewry: and this is where its also
[09:52] Zha Ewry: a little tricky
[09:52] Mojito Sorbet: The server can have its own rules about what response to give.
[09:52] Zha Ewry: So, the resource you asked for isn't moving
[09:52] Zha Ewry: but cap grant isn't exactly http get
[09:52] Mojito Sorbet: Different clients might get directed to different places, or even the same client on a different day
[09:52] Zha Ewry: which is why its al ittle odd
[09:52] Zha Ewry: 300
[09:52] Zha Ewry: says
[09:52] Zha Ewry: when you do http get/post again to this resource
[09:53] Zha Ewry: use this URI
[09:53] Mojito Sorbet: I said it is LIKE HTTP. It is a kind of redirection. I dont mean to imply anyting else
[09:53] Zha Ewry: cap grant is sent over http
[09:53] Zha Ewry: but
[09:53] Zha Ewry: its more
[09:53] Zha Ewry: Hey, SIM, I want to do X,, give me an endpoint"
[09:53] Mojito Sorbet: Forget HTTP. It is not part of this discussion
[09:53] Zha Ewry: and.. there is no coupling between the SIM URI, and the response
[09:53] Zha Ewry: k
[09:53] Zha Ewry: so, yes, its effectivel a redirect
[09:54] Zha Ewry: (I just want to not make anyone think its a redirect of the http get which is he cap reuest)
[09:54] Zha Ewry: *the
[09:54] Mojito Sorbet: By some means, over some protocol, client asks server "I want to do X". The server casn respond, "To do X, go to URI=yyy"
[09:54] Zha Ewry: Exactly
[09:54] Morgaine Dinova: I'm uncomfortable with HTTP response codes being bandied about. As Mead said very insightfully, we're guilty of webifying all this, when it should be transport-agnostic.
[09:54] Zha Ewry: and therei's added fun
[09:54] Mojito Sorbet: IT WAS AN EXAMPLE!!!
[09:54] Zha Ewry: thus my being very clear its nor, Morgaine
[09:54] Zha Ewry: we're all ont he same page
[09:54] Zha Ewry: a 300 in respose to the get
[09:54] Zha Ewry: would be to refirect the cap granter resource
[09:54] Zha Ewry: not the responde
[09:55] Zha Ewry: so.. Right, not webish
[09:55] Zha Ewry: The only bit of really annoying fun
[09:55] Zha Ewry: thats' not covered by the get side
[09:55] Zha Ewry: is why I did client side caps to begin with
[09:55] Mojito Sorbet: Some codes mean "It is over there, and don't ask me again", others mean "It ios over there right now, for you, but ask again next time"
[09:55] Dzonatas Sol: http response codes are a standard that has weight even if the codes are used elsewhere beside the transport
[09:55] Zha Ewry: The loosely coupled ends are messy
[09:55] Zha Ewry: The client is asking for "I want thevoice chat endopint"
[09:55] Mojito Sorbet: I would not even use the same numbers. The semantics will be a little different
[09:55] Mojito Sorbet: IT WAS AN EXAMPLE
[09:56] Dahlia Trimble: lol
[09:56] Morgaine Dinova: Mojito: we don't want to couple ourselves to HTTP codes.
[09:56] Zha Ewry: And the server wans to give the client the best possible endpoint
[09:56] Morgaine Dinova: kk
[09:56] Zha Ewry: which means it needs to know what the client can speak)
[09:56] Zha Ewry: and along the way, we also need to know what bits of security need to be setup
[09:56] Zha Ewry: This as morgaine observes
[09:56] Mojito Sorbet: SIP has a way of doing that. Maybe we should look at that for exmaples too
[09:56] Zha Ewry: can be wickedly expensive
[09:56] Zha Ewry: because we do it infrequently
[09:56] Zha Ewry: But, it also eneds to be wickedly flexible
[09:56] Mojito Sorbet: CLient says, "I want to do X, and I support technologies, A,B,C"
[09:57] Zha Ewry: because as soon as we get it coded in RTP"
[09:57] Zha Ewry: soemoen will come along and want it in XTP
[09:57] Zha Ewry: So, right, one way is what Mojito just suggested
[09:57] Morgaine Dinova: Well we're not going out of our way to make it expensive to set up. But we want flexibility, which usually means some extra overhead.
[09:57] Mojito Sorbet: It is only extra overhead on the session startup
[09:58] Zha Ewry: "I want voice chat cap, and I speak SIP, Vivox, and Proporarayr voice tech 78"
[09:58] Morgaine Dinova: Actually pre-session.
[09:58] Zha Ewry: and the cap granter goes
[09:58] Zha Ewry: Hmmm.
[09:58] Zha Ewry: whispers: K, "How about you use "ur://vivox cap" and "here's a key"
[09:58] Mojito Sorbet: SIP is just a wrapper. Vivox uses SIP as well. It is the codecs that vary, and that is what the AB,C is all about
[09:58] Zha Ewry: right, but vivox, is more restrictive
[09:59] Zha Ewry: or, pick another voice tool
[09:59] Latif Khalifa: says who? ;)
[09:59] Zha Ewry: the point being, the client, as it asks hints
[09:59] Zha Ewry: "I know how to talk these N ways"
[09:59] Zha Ewry: and the service goes
[09:59] Zha Ewry: Hmm. Ok, best for me, is "X"
[09:59] Zha Ewry: Client side caps went one step more indirect
[09:59] Mojito Sorbet: You can do it all in once round trip
[10:00] Zha Ewry: Defered the pipe negotation, but its not clear you need to do that
[10:00] Zha Ewry: this is probably sufficient
[10:00] Zha Ewry: The model I have in my head is:
[10:00] Dzonatas Sol: I have code for client-side caps.
[10:00] Mojito Sorbet: I want function X, and support A,B,C. Answer comes back, "from your list I choose "B". Please go to URI UUU'
[10:00] Dzonatas Sol: in use
[10:00] Zha Ewry: The fixed bits are URI (mostly URLs) which associate a bit of virtual space with a cap grant
[10:01] Zha Ewry: Everything else is subject to
[10:01] Zha Ewry: Client and Service proider pattern like this
[10:01] Zha Ewry: so.. the services are 90% facades
[10:01] Zha Ewry: You may implement them all in boring HTTP get/response an long pole on a single box
[10:01] Zha Ewry: or you can have nothign but a cap granter
[10:01] Zha Ewry: and have every single cap grant point off elsewhere
[10:02] Zha Ewry: All the external interface is in the facade
[10:02] Thoth Jantzen: maybe i missed it, or it's a dumb question, but how's this negotiation happening in the first place? what sorta connection?
[10:02] Morgaine Dinova: Not always facades. If the URI returned is scheme://IP:port/, it's not much of a facade :P
[10:03] Zha Ewry: Its still a facade, just one to one and onto in that case ;-)
[10:03] Zha Ewry: But fromt he client's perspective it knows the pattern is always
[10:03] Zha Ewry: Go to URI of resource wanted (sim in this case)
[10:03] Rex Cronon: i theory it doesn't matter the connection
[10:03] Morgaine Dinova: OK, fair enough. It's a facade in the minimalist sense that the nature of the URI is subject to interpretation at least :-)
[10:03] Zha Ewry: and then, the servcies will be granted on any pipe the client/server pair chose
[10:04] Zha Ewry: so, it may be a single monolithic process with only one type of endpoint
[10:04] Zha Ewry: or it may be 100 services sctattered in a cloud
[10:04] Zha Ewry: The client only need know how to setup the pipes
[10:04] Zha Ewry: and the services are free to be deployed as strangely as the service provdiers chose
[10:04] Zha Ewry: With one major qualifier
[10:05] Morgaine Dinova: I actually think the IP:port case is not a special case of the general service URI, and it's kind of misrepresenting it as if it were. The pattern is different. The control point is at the cap, not at the service.
[10:05] Zha Ewry: Which is its gonna be wickedly hard to seperate out services which WRAP doesn't define as seperate endpoints onto seperate service handlers.
[10:06] Zha Ewry: The pattgern, tho, is always 'Get seedcap from well known place, follwinf web rules"
[10:06] Zha Ewry: reuest caps using http get/post semantics
[10:06] Zha Ewry: and then..
[10:06] Zha Ewry: Its up to the result of the get to define how you proceed
[10:06] Zha Ewry: you *could*
[10:06] Zha Ewry: say
[10:06] Zha Ewry: The seedcap can be anything you want
[10:06] Mojito Sorbet: Not sure about "well known place". A Virtual World simulator can always answer "I don't know anyting about that"
[10:07] Morgaine Dinova: Well ... OK, not gonna argue it, since it achieves the same effect. :-) But "Cap is in control" and "Cap defers control" don't feel like the same pattern to me :-)
[10:07] Zha Ewry: The first step needs tobe wickedly concise and consistent
[10:07] Zha Ewry: You really, really, really
[10:08] Zha Ewry: don't want to do complex negotiation before the first http get
[10:08] Dzonatas Sol: that's pretty much how client-side caps works... the connection negotiation can optionally setup uris for specific caps
[10:08] Zha Ewry: But, beyond that
[10:08] Zha Ewry: the cap granter is in full control
[10:08] Dzonatas Sol: enabled/disable also
[10:08] Morgaine Dinova: I don't really get why we single out seed caps as special, since any cap can return a list of other caps.
[10:08] Zha Ewry: Because they are first opint of contant
[10:08] Mojito Sorbet: I shudder that the though of a hierarchial nae registry for the capabilities.
[10:09] Zha Ewry: You really want first point of contact to be very minmial so that the really simplistic clients don't need to bear the cost of the complex ones
[10:09] Zha Ewry: Some services may say "no, no, no" to that client a lot
[10:10] Zha Ewry: But it should be able to ask for all the starting bits in a very simple way.
[10:10] Dzonatas Sol: the registry is a simple dictionary list lookup
[10:10] Dzonatas Sol: the caps are named key, the uris are the the value
[10:10] Zha Ewry: Think of it as basically saying
[10:10] Morgaine Dinova: Mojito: hierarchy yes (because one cap yields several, recursively, it's a DAG of caps), but it's hard to avoid name registry, since we only know what caps mean by name.
[10:10] Rex Cronon: dag?
[10:10] Zha Ewry: Region gets pretty well known URL (URI for fomality, but it is oneo f a very limited set of bindings)
[10:11] Zha Ewry: from which you can get a seedcap, and then
[10:11] Tammy Nowotny: hello everyone
[10:11] Zha Ewry: Its negotiated caps all the way down
[10:11] Latif Khalifa: hi
[10:11] Rex Cronon: hi tammy
[10:11] Mojito Sorbet: I just do not want service names creeping in there unnecessarily. Like how all those fundamental JAVA classes have "sun.com" in their names
[10:11] Honour McMillan: HI :)
[10:11] Zha Ewry: heh
[10:11] Rex Cronon: what is DAG?
[10:11] Morgaine Dinova: Rex: you can think of it as a tree, in which no child can point to a parent cap or else you get cycles
[10:11] Zha Ewry: No, it would be really good to have the cap names avoid that as much as possible.
[10:11] Morgaine Dinova: Directed Acyclic Graph
[10:12] Thoth Jantzen: speechless DAWG, then. ;o)
[10:12] Dzonatas Sol: i suggest to default to cap discovery except in cases of optimization
[10:12] Zha Ewry: Strictly speaking, I'm ok if its "org.ietf.vrwap.generic_text_chat_cap"
[10:12] Zha Ewry: tho, the "org.ietf.vrwap" adds very little.
[10:13] Rex Cronon: i kow what that is, but i wasn't sure what u meant when u used that acronym:)
[10:13] Morgaine Dinova: Zha: what the prefix adds is a registry name. That's important, it's the source of capname uniqueness, because you can only have uniqueness within one organization.
[10:14] Zha Ewry: yeah, which is why I'm suggesting it exactly that way
[10:14] Zha Ewry: and by going to
[10:14] Zha Ewry: "org. ieft.vwrap"
[10:14] Zha Ewry: as the prefix
[10:14] Morgaine Dinova: Because we sure as hell don't want IANA allocating our capnames :P
[10:14] Zha Ewry: we let the DNS folks manage uniqueness
[10:14] Saijanai Kuhn: sorry about the microphone being on
[10:14] Zha Ewry: /me laughs
[10:15] Zha Ewry: "Dear hidebound organization, please allocate "temporary really cool cap for testing downloaded fabric physics paramaters"
[10:15] Morgaine Dinova: lol
[10:15] Zha Ewry: Thank you no
[10:15] Zha Ewry: I'd rather do that as
[10:15] Latif Khalifa: beyond, it would be the best if you moved from the landing point ;)
[10:15] Mojito Sorbet: PhysX claims to have fabric physics. lol
[10:15] Rex Cronon: u souldn't spell out loud your email pass, sai;)
[10:16] Beyond Baroque: Better, Latif? I have no idea how much I moved.
[10:16] Zha Ewry: "edu.fit.testwrap.fabric_phyiscis_parm_fetch"
[10:16] Saijanai Kuhn: did I say that out loud? :-/
[10:16] Zha Ewry: and know that since I am at FIT
[10:16] Zha Ewry: its my cap
[10:16] Zha Ewry: (We will for the nonce ignore versioning, please)
[10:17] Mojito Sorbet: Because of where these names get used, they will never be confused with node or class names...
[10:17] Morgaine Dinova: Caps versionsing? Eww ...
[10:17] Zha Ewry: Gonna happen, Morgaine
[10:17] Zha Ewry: soooner or later
[10:17] Morgaine Dinova: /me nods, and groans
[10:17] Zha Ewry: But
[10:17] Latif Khalifa: lol
[10:17] Rex Cronon: i was just kiddind u, i have sound off. didn't hear a thing
[10:17] Zha Ewry: I'm ok with simple string match semantics
[10:17] Zha Ewry: I have no desire to insist on
[10:17] Saijanai Kuhn: zha, wouldn't it be better to make that edu.fit.testwrap.fetch.fabric_phyiscis_parm
[10:17] Zha Ewry: "v1->v2" and some semantics about that
[10:18] Zha Ewry: heh.
[10:18] Dzonatas Sol: then you get : http://host:port/verX/articleA/verY/articleB/item
[10:18] Zha Ewry: Its opaque, Sai
[10:18] Saijanai Kuhn: ah, right full cap
[10:18] Tammy Nowotny: FIT & fabric makes me think of the university in NYC: Fashion Institute of Technology.
[10:18] Zha Ewry: Assuming that the dot structure is for anythign but human convencince is fangeorus
[10:18] Mojito Sorbet: All good protocols start off saying "This session will be at version NN.NN of the ZZZ protocol"
[10:18] Zha Ewry: That was intentional tammy ;-)
[10:18] Tammy Nowotny: :-)
[10:19] Zha Ewry: The probelm with NN.MM of ZZZ protocol
[10:19] Zha Ewry: is that you need a complete enumeration of eerything in it, for that to add much alue
[10:19] Dzonatas Sol: it assumes a set of caps
[10:19] Zha Ewry: That said, yeah
[10:19] Mojito Sorbet: Well, you could inluce a litst of tokens representing various features...
[10:19] Zha Ewry: That's the current scheme
[10:19] Tammy Nowotny: for all I know FIT may be studying virtual worlds.
[10:20] Zha Ewry: and it includes some of Zero's bias on the flavour of "anyone asking for caps will know what they want to ask"
[10:20] Mojito Sorbet: The Internet-Protocol-that-must-not-be-named uses a simple numbering, but it is a very simple protocol
[10:20] Zha Ewry: they are, Tammy.
[10:20] Tammy Nowotny: excellent!
[10:20] Zha Ewry: hetch tee tee pee!
[10:20] Mojito Sorbet: No! You have invoked the devil!
[10:20] Morgaine Dinova: Aleric made an impassioned please some 5-6 months ago to avoid the monolithic protocol versioning approach, because net-wide system can't easily evolve to support that.
[10:20] Morgaine Dinova: plea*
[10:21] Zha Ewry: and I quite agree
[10:21] Latif Khalifa: http has been 1.1 for a decade
[10:21] Morgaine Dinova: Yeah
[10:21] Zha Ewry: I want to be able to add a single cap
[10:21] Morgaine Dinova: /me nods
[10:21] Zha Ewry: or a a new style of end point
[10:21] Zha Ewry: without having to get it into version 123.78 of the protocol suite
[10:21] Mojito Sorbet: We need an update on Godwin's law. Anyone citing HTTP as a precedent ends the conversation
[10:21] Morgaine Dinova: Hahaha
[10:21] Zha Ewry: Versioning beyond describing the cap grant and response rules, ought to be per cap
[10:22] Rex Cronon: is http nazi?
[10:22] Saijanai Kuhn: Mojito's Corrolary
[10:22] Zha Ewry: So, the very high level stuff can be versioned
[10:22] Latif Khalifa: lol, so protocol version + each cap version
[10:22] Zha Ewry: as in basic patttern for initial setup
[10:22] Latif Khalifa: almost as flexible as CORBA
[10:22] Zha Ewry: but, yes, exactly Latif
[10:22] Morgaine Dinova: There should be a cocktail called Mojito's Corollary
[10:22] Zha Ewry: heh
[10:22] Mojito Sorbet: By reference, I call up the XKCD episode about WW2 generals trying to plan an attack, and being stopped by Godwin's law
[10:23] Zha Ewry: OMG they mentioned OMG tech.
[10:23] NullSubset Burner: good day all)
[10:23] Zha Ewry: /me hides under a rock making incohate noises which sound like"IOPL" and "DSOM" and rocks
[10:23] Zha Ewry: "make it stop!"
[10:23] RevMagdalen Kyrie: lol!
[10:24] Zha Ewry: I attended Object Mnagement Group meetings in a previous life
[10:24] Zha Ewry: One goal for VWRAP is to avoid most of those traps and experiences
[10:24] Latif Khalifa: did you by any chance work on SOAP? ;)
[10:24] Zha Ewry: Wash your mouth out Latif
[10:24] Latif Khalifa: :P
[10:25] Zha Ewry: SOAP looks fine, until you try and use it
[10:25] Latif Khalifa: exactly
[10:25] Latif Khalifa: overengineered POS nobody wants to use unless they have to
[10:25] Zha Ewry: at which point, you find out that ersioning objects across the web is a really bad thing
[10:25] Mojito Sorbet: Is SOAP that Service Oriented something Protocol?
[10:25] Morgaine Dinova: I wonder why CORBA failed. It's not hugely different to many other RPC systems.
[10:25] Zha Ewry: Both CORBA (OMG) stuff and SOAP had way too much of a habit of binding object semantics onto RPC
[10:25] Latif Khalifa: Mojito, SOAP is xml web services RPC system
[10:26] Mojito Sorbet: Yes, ok, I know about that.
[10:26] Zha Ewry: Too much end-to-end semantics Morgaine
[10:26] Rex Cronon: corba can be a pain 2 use:(
[10:26] Zha Ewry: Wickedly hard to version
[10:26] Gazanfer Jehangir: CORBA was meant for lehacy systems support which die on their own
[10:26] Latif Khalifa: you see, people make a comittee like wrap, people sit on them and invent all possible and impossilbe uses and edge cases
[10:26] Morgaine Dinova: Zha: you mean like we are in danger of doing, if we're not careful? :-)
[10:27] Zha Ewry: Heh
[10:27] Mojito Sorbet: SOAP is popular in B2B things, so the books claim
[10:27] Latif Khalifa: and the nobody uses that crap and they go for KISS solutions likeREST
[10:27] Zha Ewry: Thats why the end goal for 90% of WRAP is RESTful
[10:27] Morgaine Dinova: Not joking. It's why I'm focused on the low level efficiency end of the use cases atm
[10:27] Zha Ewry: also not kiding
[10:27] Zha Ewry: which is why I'm pushign back hard on any attempt to do hueg protocol suites in favor of
[10:27] Latif Khalifa: you're wrapping corba/soap carma in a restful protocol
[10:27] Latif Khalifa: karma*
[10:27] Zha Ewry: "Here is a cap for doing X"
[10:28] Zha Ewry: And, keeping the basic patterns really limited
[10:28] Latif Khalifa: basic patterns?
[10:28] Zha Ewry: VWRAP should be mostly doing two very seperaable tasks in my mind
[10:28] Zha Ewry: One is the basic framework
[10:28] Zha Ewry: (how we setup stuffs)
[10:28] Latif Khalifa: like having to check the protocol version and then get the list of caps and then ask each cap about its own version and handle all the combinations? you call that basic?
[10:29] Zha Ewry: Except you don't Latiff
[10:29] Zha Ewry: You do an http post with alist of cap strings
[10:29] Zha Ewry: and alist of client side endpoints you're willing ot use
[10:29] Zha Ewry: and you get back a set of caps
[10:29] Zha Ewry: If some of those repeat the pattern
[10:29] Zha Ewry: (which I expect will be rare)
[10:30] Zha Ewry: the code is identical
[10:30] Zha Ewry: Thus
[10:30] Mojito Sorbet: I am having trouble seeing the application for all this...
[10:30] Zha Ewry: "No versioning of the overall protocol, beyond making sure you are using the correct setup scheme
[10:30] Latif Khalifa: Mojito, I do to
[10:30] Gazanfer Jehangir: i have one in mind
[10:31] Latif Khalifa: well you can always imagine some edge case were it's absolutelly neccessary...
[10:31] Dahlia Trimble: dumb http question: why POST instead of GET? a GET URL can be quite long
[10:31] Mojito Sorbet: POST implies you are changing state on the server
[10:31] Gazanfer Jehangir: its quite usbale but again interoperable over how much and how many systems?
[10:31] Dahlia Trimble: but your not changing state
[10:31] Mojito Sorbet: GET is not supposed to change any server state (tho sometimes people cheat)
[10:31] Zha Ewry: This is setting up an association, which I think is a state change (1)
[10:32] Latif Khalifa: Dahlia, you don't want to send UploadBakedTexture with GET xD
[10:32] Morgaine Dinova: Well with the degenerate form of URI as IP:port in the cap return block, which will be very appropriate for small standalone systems, there really is no setup overhead for the streaming transport. It doesn't get leaner.
[10:32] Dahlia Trimble: I mean for Capability requests
[10:32] Zha Ewry: and 2) I prefer to push things like pais of lists of strings as a body with apost, not a get
[10:32] Latif Khalifa: Dahlia, UploadBakedTexture is a cap
[10:32] Zha Ewry: *pairs
[10:33] Mojito Sorbet: Yes, a POST with a bunch of JSON in the body
[10:33] Morgaine Dinova: Arghh. No HTTP words! You mean one of the 3 CRUD operations.
[10:33] Zha Ewry: I think its going to be a post of a fairly big wodge of strings
[10:33] Morgaine Dinova: 4
[10:33] Zha Ewry: Well
[10:34] Zha Ewry: On the very first cap grant, calls, I think it really is http
[10:34] Dzonatas Sol: How you build C++ is basically SOAP
[10:34] Zha Ewry: byeond that, tho, I'm totally fine calling it CRUD
[10:34] Dzonatas Sol: C++ objects*
[10:35] Rex Cronon: how can u compare c
[10:35] Tammy Nowotny: I havent heard this song in eons
[10:35] Rex Cronon: ++ and soap?
[10:35] Saijanai Kuhn: seaside webservers tack on a session ID and page id to the end of the URL. That might make GET commands shorter than you might like: http://obi.ath.cx:8080/examples/playground?_s=sq1ns2MvQnASyIDl&_k=W81HhWG3pmThki8t
[10:36] Latif Khalifa: lol
[10:36] Zha Ewry: In terms of post, its also wickedly easier to build up a body than a URL/URI
[10:36] Dzonatas Sol: SOAP is simple object access protocol... and the way you expect to "message" c++ is the basics of SOAP
[10:36] Morgaine Dinova: Sai: but Seaside doesn't pretend to be transport-agnostic. We are, so no webbifying pls :P
[10:37] Zha Ewry: and if you ever want to "post" onto other endpoints
[10:37] Zha Ewry: keeping the content in the body, is much cleaner
[10:37] Saijanai Kuhn: well, get in the URL is certainly more webified than constructing a POST body
[10:37] Techwolf Lupindo: I'me not familar with SOAP, but you use it durrning a shower....*ducks*....but anyway, I've read where some large projects got away from SOAP.
[10:37] Zha Ewry: masively so
[10:37] Dzonatas Sol: UML is still tied heavily to SOAP
[10:38] Zha Ewry: I don't think anyone who's built stuff on SOAP isgoing to push it here ;)
[10:38] Latif Khalifa: nope
[10:38] Latif Khalifa: especially using extensions like ws.security
[10:38] Latif Khalifa: /me makes a hand sign to ward off evil :P
[10:39] Rex Cronon: soap is a protocol, not a programming language
[10:39] Zha Ewry: /me groans at ws.*
[10:39] Latif Khalifa: i have the scars to show for it ;)
[10:39] Zha Ewry: Morgaine, have you looked at encoding your comments to meah, in a set of proposed changes to the current draft?
[10:40] Dzonatas Sol: how you access an object is protocol
[10:40] Dzonatas Sol: they call it SOAP instead of SOA two distinguish it from Sevice Oriented Architecture
[10:41] Dzonatas Sol: too*
[10:41] Latif Khalifa: SOAP is a very specific xml-over-http rpc protocol, nothing to do with SOA except a fancy name
[10:42] Tammy Nowotny: I am being called away on business
[10:42] Tammy Nowotny: TY everyone
[10:42] Dzonatas Sol: SOAP doesn't even demand xml or http... that is just a common form you find
[10:42] Morgaine Dinova: Zha: she's currently working on that herself, although to be truthful, I' a bit uncertain what she has in mind. I liked her basic idea to tidy up LLIDL (I'm hoping we can clean up the syntax as a side effect of losing HTTP references), and I liked the opportunity to add some solutions for how we kick off special transports like for update streams.
[10:42] Rex Cronon: tc tammy
[10:42] Zha Ewry: k
[10:43] Latif Khalifa: llidl is another reason vwrap smells more and more like soap (and it's wsdl)
[10:43] Dzonatas Sol: SOAP allows ambiguous object access methods... and REST narrows it down to 4 methods
[10:43] Zha Ewry: I'd like to make sure we get a very clean versino of this discussion summed up and back int he specs
[10:43] Dzonatas Sol: for example
[10:44] Dzonatas Sol: try to make a REST model in UML and you'll understand SOAP
[10:44] Zha Ewry: and never code in it ;)
[10:44] Morgaine Dinova: Well LLIDL is the reason why AWG never actually created that list of REST nouns that I asked Zha for countless times since the founding of the group. :-) I think LLIDL is shunned :P
[10:45] Zha Ewry: LLIDL tends to come awfully close to a fetish of Zeros. Nobody codes or thinks in it.
[10:45] Morgaine Dinova: Aye. but its main problem is the stupid markup. people don't like it for the same (silly) reason they don't like LISP: too many funny characters :P
[10:46] Dzonatas Sol: i just find I have to reference the symbols to understand the same print each time i look at llidl
[10:46] NullSubset Burner: Soooooooo
[10:46] NullSubset Burner: might be mistaken but didnt SOA start as a C-extensiable libray- realtes closley in terms of untility/useage?
[10:47] Morgaine Dinova: If I understood Mead correctly (and I rather fear that I didn't), she wants to turn LLIDL into a pure C-like structure tree, with resources separate from abstract data types.
[10:47] Morgaine Dinova: And with no HTTP'ish markup.
[10:49] Dzonatas Sol: almost 11....
[10:49] Morgaine Dinova: Which would be very clean.
[10:49] Dzonatas Sol: take care =)
[10:49] Morgaine Dinova: And using the 4 CRUD operator words explicitly to identify resource semantics.
[10:49] Rex Cronon: r u still here zha?
[10:50] Rex Cronon: there is one "little" thing that needs to be addressed
[10:50] Zha Ewry: I am
[10:50] Zha Ewry: I was off woolgathering on REST nouns for s moment
[10:50] Rex Cronon: our groups charter needs to be updated
[10:51] Rex Cronon: if u read it u will see what i am talkingabout
[10:51] Zha Ewry: What would you like changed Rex? I haven't actually looked in ages
[10:51] Sridhar Shepherd: where is current charter
[10:51] Beyond Baroque: "sponsored by LL"?
[10:52] Zha Ewry: Yeah, pretty backlevel
[10:52] Morgaine Dinova: Fair enough, Rex is right, it's out of date in various areas.
[10:52] Rex Cronon: click on group info in group chat
[10:52] Morgaine Dinova: Basically, remove the LL angle, and add the VWRAP one.
[10:52] Rex Cronon: don't need to change it now, but keep it in mind
[10:52] Zha Ewry: And probably add a warning that you will be exposed to odd techie chat at 1:00 am SL time
[10:53] Zha Ewry: I will pull together an update and post it to the group for comments
[10:53] Morgaine Dinova: You already added that Zha, or another officer did
[10:53] Latif Khalifa: 24/7
[10:54] Rex Cronon: oh. it might be a good idea to post a LM to this place, and when and how long do these group meetings take place
[10:54] Zha Ewry: It wanrs you words may travel far, but not so much that at 2:00 am you may get a four page debate on the efficicact of fusion power generations schemes
[10:54] Dahlia Trimble: I always LOL when people join a chat group and then complain that people are using it to chat
[10:54] NullSubset Burner: lol
[10:54] Zha Ewry: Yeh, those would be goodt hings
[10:54] Zha Ewry: I'll make sure they get in
[10:54] Morgaine Dinova: Hehe
[10:54] NullSubset Burner: anti - garivty next:P
[10:54] Zha Ewry: So, I'll do that
[10:55] Zha Ewry: and I'll chew al ittle on whether we're at a point where nouns is a good discusson for next week
[10:55] Zha Ewry: I think it might be fun to build the list here, Morgaine
[10:55] Zha Ewry: (Its much eaiser to do such things without worrrying abotu stepping on Lidnen's toes. )
[10:55] Morgaine Dinova: You really ought to appoint another officer Zha. Tree is never around, and Sai is asleep most of the time. We had someone a few days ago asking for an officer, there were none around.
[10:55] Latif Khalifa: i want DOWHATIWANT verb :P
[10:56] Zha Ewry: You don't get verbs, only nouns ;-)
[10:56] Zha Ewry: is REST we know our verbs ;-)
[10:56] Morgaine Dinova: Only 4 verbs, the CRUD ones
[10:56] Zha Ewry: well, is mostly rest
[10:56] Zha Ewry: Anyone else want to be an officer in the zoo?
[10:56] Morgaine Dinova: Anyone who WANTS shouldn't get :-)
[10:57] Rex Cronon: i nominate morgaine:)
[10:57] Zha Ewry: Heh\
[10:57] Latif Khalifa: don't I have that? :P
[10:57] Zha Ewry: You're close Latif, you can moderate
[10:57] Latif Khalifa: to kill off rogue bots :P
[10:57] Tammy Nowotny: sorry if I stepped on anyone
[10:57] Zha Ewry: Morgaine, you're on nearly 24/7? Want invite/mute powers?
[10:57] Latif Khalifa: ah, luckily didn't have a use for chat moderation yet
[10:58] Zha Ewry: That's best ;)
[10:58] Latif Khalifa: yeah :D
[10:58] Zha Ewry: I try, wickedly hard to be light handed on all of it
[10:58] Morgaine Dinova: /me shrugs. I don't mind helping. But unlike Sai, I'm not going to invite anything that moves, and corpses as well.
[10:58] Tammy Nowotny: Morgaine is on even more than Zha & Sai
[10:58] Latif Khalifa: lol
[10:58] Rex Cronon: lol morgaine
[10:58] Zha Ewry: That's fine Morgaine
[10:58] Zha Ewry: I am once every so often
[10:58] Zha Ewry: tempted to kick out 90% of the group
[10:58] Latif Khalifa: who invited Wrapp deserves a medal lol
[10:58] Beyond Baroque: /me still wonders how he ended up in this group. That might explain it...
[10:59] Zha Ewry: and ask the lurkers to justify why they are here
[10:59] Zha Ewry: but..its way too much like work
[10:59] Zha Ewry: is wrapp still muted?
[10:59] Latif Khalifa: no
[10:59] Zha Ewry: /me tries to recall if she muted wrapp personally or group wide
[10:59] Latif Khalifa: he was only muted per session
[10:59] Zha Ewry: heh
[11:00] Morgaine Dinova: Poor Wrapp. But sure it's a medic al condition, but certainly annoying for us.
[11:00] Latif Khalifa: anyway, i'm jumping over to andrews, take care yu all
[11:00] Morgaine Dinova: I'm off to OSgrid
[11:00] NullSubset Burner: my timing is apparently impcable as usual:- 90% kick:(-lol
[11:00] Rex Cronon: does wrapp expect a medal for leaving his/hers bot going bla, bla, bla?
[11:00] Dahlia Trimble: I'm outta here too, bye all :)
[11:00] Rex Cronon: going to andrew's oh too
[11:00] Morgaine Dinova: Cya Dahlia, Rex, et all
[11:00] Rex Cronon: tc everybody
[11:01] Saijanai Kuhn: will have transcript kup later today
[11:01] Beyond Baroque: /mw waves and poofs
[11:01] Tammy Nowotny: tc, back to RL-grid myself
[11:01] Morgaine Dinova: OMG, hi Sai ... right after I took your name in vain :P
[11:01] Tammy Nowotny: I had a funny experience: my Rl self ran into Tuna Ofddfellow's at a talk at the Harvard law School
[11:02] Tammy Nowotny: I eended it by saying I was going back to Real life...when of course this already was RL
[11:02] RevMagdalen Kyrie: lol!
[11:02] Tammy Nowotny: but he said "Harvard is not real life" which is true
[11:02] Zha Ewry: For epislons around real
[11:03] NullSubset Burner: Nice to meet you folks take care-(hope to attend again!!)
[11:03] Tammy Nowotny: anyway, see you all soon
[11:03] RevMagdalen Kyrie: see you guys!
[11:03] Zha Ewry: Eeep
[11:03] Zha Ewry: Off to next RL thing
[11:03] Tammy Nowotny: :D
[11:03] Zha Ewry: I'll ponder chater
[11:03] Tammy Nowotny: *poof*
[11:03] Zha Ewry: and expect nouns on tap for next week
[11:05] Morgaine Dinova: Off to Andrew's on this client, but AFK as in OSgrid
[11:07] Thoth Jantzen: hmm
[11:07] Thoth Jantzen: did i miss the end??
[11:07] Thoth Jantzen: ;o)

AW Groupies/Chat Logs/AWGroupies-2010-07-27

Navigation menu

Search