#rest on irc.freenode.net 2010-05-04 (txt)

[00:52] bigbluehat (~byoung@adsl-177-117-159.gsp.bellsouth.net) left irc: Quit: bigbluehat
[00:53] KevBurnsJr (~kevburnsj@c-67-161-65-81.hsd1.ca.comcast.net) left irc:
[01:39] avdi (~avdi@c-174-55-88-101.hsd1.pa.comcast.net) joined #rest.
[02:45] avdi (~avdi@c-174-55-88-101.hsd1.pa.comcast.net) left irc: Quit: avdi
[03:00] RestBot joined #rest.
[06:18] algermissen (~algermiss@p54907F83.dip.t-dialin.net) joined #rest.
[06:20] algermissen (~algermiss@p54907F83.dip.t-dialin.net) left irc: Client Quit
[07:01] KarlHungus (ssanders@unaffiliated/adj) left irc: Ping timeout: 240 seconds
[07:03] KarlHungus (ssanders@for.my.n1gs.net) joined #rest.
[07:08] algermissen (~algermiss@tmo-100-178.customers.d1-online.com) joined #rest.
[07:42] mhausenblas (~mhausenbl@wg1-nat.fwgal01.deri.ie) joined #rest.
[07:51] algermissen (~algermiss@tmo-100-178.customers.d1-online.com) left irc: Quit: RESTing
[07:57] algermissen (~algermiss@tmo-100-32.customers.d1-online.com) joined #rest.
[08:14] ordnungswidrig (~Adium@91.208.107.57) joined #rest.
[08:16] ordnungswidrig (~Adium@91.208.107.57) left irc: Client Quit
[08:24] ordnungswidrig (~Adium@91.208.107.57) joined #rest.
[08:59] algermissen (~algermiss@tmo-100-32.customers.d1-online.com) left irc: Ping timeout: 240 seconds
[09:40] algermissen (~algermiss@p54907F83.dip.t-dialin.net) joined #rest.
[10:12] algermissen (~algermiss@p54907F83.dip.t-dialin.net) left irc: Quit: RESTing
[10:25] algermissen (~algermiss@p54907F83.dip.t-dialin.net) joined #rest.
[10:26] algermissen (~algermiss@p54907F83.dip.t-dialin.net) left irc: Client Quit
[10:36] algermissen (~algermiss@p54907F83.dip.t-dialin.net) joined #rest.
[12:56] bradley-holt (~bradley-h@65-183-135-35-dhcp.burlingtontelecom.net) joined #rest.
[12:56] <darrelmiller> I may regret this but I decided to get feisty with the Rails guys on Stackoverflow. I'm sick of hearing, "oh your service is more than just CRUD, REST doesn't handle that very well"
[12:57] <darrelmiller> Oh and if you want another laugh. Check out the problems this library is supposed to be solving http://hammock.codeplex.com/wikipage?title=Futures
[12:58] <darrelmiller> This is a classic "Unfortunately, APIs change so frequently and proliferate constantly, making it tedious, time-consuming, and error-prone to maintain wrapper code"
[13:11] <mikekelly> ://
[13:11] <mikekelly> link me to the thread where you unload on the rails peeps
[13:13] <mikekelly> to be fair - that kind of sentiment isn't reserved to rails
[13:22] <darrelmiller> http://stackoverflow.com/questions/2764814/restful-membership
[13:32] <darrelmiller> mikekelly: You are absolutely right, it is not just the Rails guys who make this assertion. If you think my comments are out of line, I'll revise them. I was just annoyed by the cavalier attitude.
[13:44] ordnungswidrig (~Adium@91.208.107.57) left irc: Ping timeout: 240 seconds
[13:50] mamund checks in
[13:57] <darrelmiller> Morning mamund. This working on real stuff sucks. Can we have another conference please?
[14:00] <mamund> hehehe
[14:00] <mamund> i know what you mean...
[14:00] <mamund> been swamped since my return
[14:01] darrelmiller has to leave and actually go and visit a customer.
[14:21] bigbluehat (~byoung@adsl-71-189-114.gsp.bellsouth.net) joined #rest.
[14:57] <mikekelly> the thing about rails is that you can actually over-ride all the routes
[14:58] <mikekelly> so there's nothing actually stopping the 'Resource Controller' pattern you're detailing there
[14:58] <mikekelly> it's just that the default rails 'DSL' for resources is based on the collection+item thing
[14:59] <mikekelly> the main value in rails is all the ORM stuff i.e. activerecord
[15:00] <mikekelly> but yeah - they use 'resource' in the DRM where they really mean 'collection'
[15:00] <mikekelly> DRM? I meant DSL :]
[15:03] <mikekelly> sinatra is a way better approach imo
[15:04] <mikekelly> the problem is that rails has a lot of stuff that'll bolt in 'for free'
[15:04] <mikekelly> but now they're moving across to rack, hopefully most of those components will end up being easily workable in sinatra
[15:09] <mikekelly> sinatra/rack is the best web app "framework" I've seen
[15:09] <mikekelly> by a mile
[15:15] <mikekelly> you know it's pure win when hello world looks like this:
[15:15] <mikekelly> require 'sinatra'
[15:15] <mikekelly> get '/hi' do "Hello World!"
[15:15] <mikekelly> end
[15:16] <mikekelly> hmm "Hello World!" should've been on a new line :(
[15:30] darrelmiller (~darrelmil@bas4-montreal45-1242355736.dsl.bell.ca) left irc: Ping timeout: 265 seconds
[17:13] mhausenblas (~mhausenbl@wg1-nat.fwgal01.deri.ie) left irc: Quit: mhausenblas
[17:47] ordnungswidrig1 (~Adium@p5B173E4E.dip.t-dialin.net) joined #rest.
[18:03] bigbluehat (~byoung@adsl-71-189-114.gsp.bellsouth.net) left irc: Quit: bigbluehat
[18:06] ordnungswidrig1 -> ordnungswidrog
[18:16] bradley-holt (~bradley-h@65-183-135-35-dhcp.burlingtontelecom.net) left irc: Remote host closed the connection
[18:18] <mikekelly> jesus christ
[18:18] <mikekelly> "information resources"
[18:18] <mikekelly> bloody hell.
[18:19] <mikekelly> I mean - seriously
[18:19] <jsled> looking at range-14 stuff, or ?
[18:19] <mikekelly> cmon..
[18:19] <mikekelly> yeah again
[18:19] <jsled> .g "information resources"
[18:19] <phenny> jsled: http://www.symphonyiri.com/
[18:19] <mikekelly> whoever brought up that as a 'problem' is insane
[18:20] <mikekelly> ok so I start building some system on the web that is so crap I ahve to make stupid arbitrary distinctions between supposedly 'uniform' identifiers
[18:20] <mikekelly> that is all kinds of dumb.
[18:23] <mikekelly> aoh it's a timbl thing
[18:23] <mikekelly> I didn't realise it was him who brought that up
[18:23] <mamund> it's all tied up w/ his Semantic Web thing
[18:23] <mikekelly> well whatever, call me a heathen but I personally think it's bullshit and if it's important than the system he's trying to build is shit
[18:23] <mamund> actually, tied up in inter-document URIs.
[18:24] <mamund> yeah.
[18:24] <mamund> it's the biggest single thing crippling this whole "SemWeb" good-ness
[18:25] <jsled> really?
[18:25] <mikekelly> the thing is.. there already is a mechanism for showing documents that are representations of a resource and also have their own URI
[18:25] <mikekelly> it's called conneg, var, and content-location
[18:25] <mikekelly> vary^
[18:27] <mikekelly> I don't know whether semweb has any claim to 'triples' but that's all I care about personally
[18:27] <mikekelly> they can go play with all their semantics and sit rubbing their hands waiting for artificial conciousness to arrive
[18:27] <mikekelly> I'll go build stuff that people actually use.
[18:28] <mamund> i've been 'round this a couple times and it never comes out well...
[18:28] <jsled> "stuff people will actually use" is probably disjoint from "triples", though.
[18:29] <mikekelly> when I say people I mean developers programming machiens
[18:29] <jsled> sure.
[18:29] <ordnungswidrog> jsled: you can hide the triples-stuff from the poople and let it be an implementation detail
[18:29] <mikekelly> if it's people people I'd use html..
[18:29] <mikekelly> html is kind of triples
[18:29] <mamund> the desire to make assertions based soley on the URI is just never going to work out well.
[18:29] <ordnungswidrog> mikekelly: there are enough developers that don't get triples, right?
[18:29] <mikekelly> just very weird triples that get to use images and stuff to imply the link relation
[18:29] <jsled> If that's the case, then what's the use of triples? And, I'd argue, you can't effectively do so if you insist on triples.
[18:30] <mikekelly> I dont' think triples are all that complicated
[18:30] <mikekelly> triples are just hte basic essence of maachine hypertext
[18:30] <mikekelly> infact, that's all machine hypertext is.
[18:31] <ordnungswidrog> mikekelly: but that is a very low level abstraction.
[18:31] <mikekelly> ? it's what developers will use
[18:33] <ordnungswidrog> yes, but you will have do define concepts like "collections" or "types" on top of them, right? Else it's like saying the essence is a stream of 1s and 0s
[18:33] <mikekelly> yeah I'll have to define them in my documentation
[18:34] <mikekelly> you know.. the bit that developers read when they program a client :P
[18:34] <jsled> Or "photos" or "likes".
[18:35] <ordnungswidrog> developers read documentation? My experience say, you must inject documentation using a big fat cannon or like that :-)
[18:35] <mikekelly> if your link relations resolve
[18:36] <ordnungswidrog> if your client knows the semantic
[18:36] <mikekelly> to some html page that says - hey dude expect this kind of thing here .. *example(*
[18:36] <mikekelly> maybe even drops them into your documentation app
[18:36] <mikekelly> :)
[18:37] <ordnungswidrog> doc app, great idea
[18:38] <mikekelly> then we're back to "wouldn't it be cool if your media types were identified by URIs" thing
[18:38] <mikekelly> :)
[18:38] <ordnungswidrog> if everbody could agree on doctypes for that. docbook, ok, javadoc, ok, but there is such a big mess of documentation types...
[18:38] <mikekelly> html?
[18:38] <ordnungswidrog> mikekelly: too generic
[18:39] <ordnungswidrog> mikekelly: at least html enriched wich ctags or like that.
[18:39] <mikekelly> html if you wanted to formalise it
[18:39] <mikekelly> you're talking person to person
[18:39] <ordnungswidrog> text/javadoc+html
[18:40] <mikekelly> if you wanted to formalise it you might create a media type
[18:40] <ordnungswidrog> I, as a developer, want to lookup a symbol from a document resource, right?
[18:40] <mikekelly> as a developer you should care most about what each link relations mean
[18:43] <ordnungswidrog> I'd enjoy if we could come back to a more decentral internet, currently it's a facebook-google-yahoo-youtube-hub not a web.
[18:44] <mikekelly> that's an economic and social thing as much as technical
[18:44] <mikekelly> probably necessary in evolutionary terms
[18:44] <mikekelly> but they are dinosaurs
[18:44] <mikekelly> dinosaurs will die
[18:44] <ordnungswidrog> it's interesting that facebook goes into aRDF and opengraph will beeing the largest social web hub.
[18:45] <ordnungswidrog> I think they might turn from being a content provider to a search engine.
[18:47] <mikekelly> what tv is now
[18:47] <mikekelly> that's what they'll be the equivalent of in 5 years time
[18:47] <ordnungswidrog> or what the village's market places used to be.
[18:48] <ordnungswidrog> reminds me of talk during school breaks
[18:48] <mikekelly> it's a cost.. someone has to find a way to pay for us to have our little facebook web pages
[18:48] bigbluehat (~byoung@adsl-177-117-159.gsp.bellsouth.net) joined #rest.
[18:49] <ordnungswidrog> mikekelly: blogger, tumblr, there are free blogs, e.g.
[18:50] <mikekelly> yeah there are - but there are no decent tools that handle the social interconnectivity with the same level of convenience as facebook
[18:51] <ordnungswidrog> right. But I think this is where facebook is heading to. (Wild guess).
[18:51] <mikekelly> you can get the fairest news by scouring the web for topics.. but people still read newspapers and watch news programs
[18:52] <ordnungswidrog> yes, it will only work for news of your social web, like facebook / twitter etc.
[18:53] ordnungswidrog -> ordnungswidrig
[18:53] <mikekelly> still centralised broadcasting though
[18:53] <mikekelly> you're at the mercy of someone else's programming
[18:53] <mikekelly> different kind of programming than what you get on tv of course :)
[18:55] <mikekelly> I wonder how long it will be until google is a SPARQL endpoint?
[18:55] <mikekelly> :P
[18:58] <ordnungswidrig> I think facebook will be first. :-P
[18:59] <ordnungswidrig> Is there a sparql search engine?
[19:00] <mikekelly> dunno much about that stuff I don't really use or like facebook
[19:04] <mikekelly> mamund: asking semweb guys what this means - "HTTP URIs define a web of information objects"
[19:04] <mikekelly> why define instead of describe
[19:04] <mikekelly> and what does objects actually mean and why is it necessary
[19:05] <ordnungswidrig> the uris define a web by describing information objects
[19:08] <mamund> hehehe....
[19:08] <mamund> ask 'em yourself, dude!
[19:08] <mikekelly> yeah i just did no-one's bitten
[19:08] <mamund> join #swig here on freenode
[19:08] <mikekelly> already there :P
[19:08] <mamund> dig in
[19:08] <mamund> mhausenblas will take up for you.
[19:09] <mamund> to carry this a bit farther...
[19:09] <mikekelly> good point
[19:09] <mikekelly> :D
[19:09] <mamund> the semweb folks think in terms of classes and objects exposed on the web
[19:09] <mamund> RDF is not a representation for them, it's the class/object
[19:10] <mamund> makes things tough for widely dist systems work
[19:10] <mikekelly> right - that's ok because RDF isn't a serialization it's just a framework
[19:10] <mikekelly> RDF/xml is a representation
[19:10] <mamund> yeah, but right now RDF/XML just serializes the class/object data
[19:11] <mamund> same for n3
[19:11] <mikekelly> that's alright too it's still a representation of the resource though right?
[19:12] <mamund> it represents the object, but none of the network semantics...
[19:12] <mamund> and, so far, represents the *whole* object...
[19:12] <mamund> which makes representation passing difficult
[19:12] <mikekelly> say what now
[19:12] <mamund> if my FOAF file is really big....
[19:13] <mamund> there are now clear ways to send just part of it to an agent, have the agent edit and return that to a server
[19:13] <mamund> the entier FOAF is the only reasonable representation
[19:14] <mikekelly> what PATCH?
[19:14] <mikekelly> I think I may be super confused right now
[19:14] <mamund> PATCH could work if there was a media type for it
[19:14] <mamund> (appliation/rdf-patch)
[19:16] <mikekelly> ok so problem solved?
[19:17] <mamund> maybe
[19:17] <ordnungswidrig> REST pattern #443: "If in doubt: throw a media type at it"
[19:17] <mikekelly> ;/
[19:17] <mamund> ordnungswidrig: bingo! (hahah)
[19:20] mamund skips over to #swig to watch mikekelly in action...
[19:25] <mikekelly> ok that guy is just confusing me
[19:39] <mikekelly> my problem is this
[19:39] <mikekelly> http://www.w3.org/DesignIssues/HTTP-URI#identifies
[19:39] <mamund> thas she blows!
[19:39] <mikekelly> I don't agree with that
[19:39] <mamund> s/thas/thar
[19:40] <mikekelly> "the problem is we didn't realise that would be a good idea and have created a crap system that's made us have to care about this nonsense"
[19:41] <mikekelly> bleh I think I'm giving up on this already
[19:42] <mikekelly> some people think we're robots, or that robots are us - and so they'll continually inflict their anxiety ridden explanations of the world on us
[19:42] <mamund> that's what most do. hence the "solution" is to use the response code and be done w/ it.
[19:42] <ordnungswidrig> I don't get it. A person can have multiple home pages, as well as a web site can be about multiple persons
[19:42] <mamund> you sound like Alan Watts!
[19:43] <mikekelly> yeah apart from alan watt's would be nice about it
[19:43] <mikekelly> :P
[19:43] <mikekelly> don't know what that apostraphe was for
[19:44] <ordnungswidrig> mikekelly: spare apostrophe
[19:44] <mamund> http://en.wikipedia.org/wiki/The_map_is_not_the_territory <- love this guy
[19:45] <mikekelly> ce n'est pas un pipe or whatever
[19:45] <mikekelly> I might do an artists interpretation of REST vs semweb/linkeddata
[19:46] <mamund> i actually thought SemWeb should have used a new scheme (sw://) to help solve this problem
[19:46] <mikekelly> rest side will say "the map is not the territory"
[19:46] <mikekelly> and the semweb side will say "..... fuck you we've got tim berners lee on our side"
[19:47] <ordnungswidrig> so try to grow some corn on you map, then.
[19:47] <mikekelly> haha
[19:47] <mikekelly> no no semweb side should say "not yet it's not"
[19:48] <mikekelly> and they should be furiourlsy scribbling on bits of paper and going mad
[19:50] <mikekelly> "the problem of making sense out of the seeming chaos of experience reminds me of my childish desire to send someone a parcel of water in the mail. The recipient unties the string, realeasing the deluge in his lap. But
[19:52] <mikekelly> the game would never work since it is irritating and impossible to wrap and tie a pound of water in a paper package. There are kinds of paper which won't disintegrate when wet, but the trouble is to get the water into any managable shape, and to tie the string without bursting the bundle"
[19:54] <mikekelly> "The more one studies attempted solutions to problems in politics and economics, in art, philosophy, and religion, the more one has the impression of extremely gifted people wearing out their ingenuity at the impossible and futile task of trying to get the water of life into neat and permanent packages"
[19:59] <mamund> i have a printed copy of this email taped to my wall right now: http://lists.w3.org/Archives/Public/www-tag/2005Jun/0039.html
[20:06] <mikekelly> hmm
[20:06] <mikekelly> timbl:
[20:06] <mikekelly> "If you refer to the thing as a dog, and your sever says 200 to an
[20:06] <mikekelly> HTTP get for it, one of you is wrong. It is an error."
[20:07] <mamund> algermissen must have loved that first response
[20:08] <mikekelly> I really don't like the sound of that timbl quote :/
[20:09] <mikekelly> is that just 'from his perspective' or 'if you do this on the web your wrong' ?
[20:12] <mamund> i have always read it as "wrong if you're following this new TAG resolution"...
[20:12] <mikekelly> ah that is it
[20:12] <mamund> IMO, it's not much diff than some Roy responses.
[20:12] <mikekelly> Roy said it perfectly
[20:12] <mikekelly> Once we accept that it is impossible to disambiguate resources
[20:12] <mikekelly> by changing their names (because people will simply use those
[20:12] <mikekelly> new names in ambiguous ways), we might be able to make progress
[20:12] <mikekelly> on the real problems of the SemWeb by applying other methods of
[20:12] <mikekelly> disambiguation.
[20:13] <mikekelly> I wish I'd have known beforehand I'd have had it printed on a T shirt for WWW2010
[20:14] <mikekelly> that is, I think, a lot of what is at the core of those pedantic web anxieties
[20:15] <ordnungswidrig> TBL's argument the HTTP URIs (without "#") should be understood as referring to documents, not cars.
[20:18] <mikekelly> this is awesome too:
[20:18] <mikekelly> "Namespace trees should not be forced into a flat topology just
[20:18] <mikekelly> because some people think an abstraction can't be represented
[20:18] <mikekelly> on the Web. Those people are wrong -- there are millions of
[20:18] <mikekelly> abstractions represented on the Web right now and they aren't
[20:18] <mikekelly> going to go away just to make it easier for computers to do AI."
[20:18] <ordnungswidrig> heh
[20:18] <ordnungswidrig> *G*
[20:20] <mamund> it's the last para of that email that really cuts to the heart of it...
[20:21] <mamund> i get the sense he has a better solution that none of the involved parties were ready to accept.
[20:22] <mikekelly> well he wants their solution to stare that problem in the face
[20:22] <mikekelly> and not try and jump over it
[20:23] <mikekelly> and/or bury it somewhere and pretend it doesn't exist
[20:26] <mamund> the good news about the resolution is that it's not URI- or scheme-based...
[20:27] <mamund> that means it's possible that Roy's link relation idea (i have that correct?) could be instituted in the future.
[20:32] <mikekelly> I really think the primary point of confusion here, and it happens a lot with the hypermedia constraint aswell
[20:32] <mikekelly> is that context is not only ok... it's actually fundamental to understanding what a resource is
[20:34] <mikekelly> so the question "should that URI identify a car or an information resource about the car" depends on the context that you approach it from
[20:35] <mikekelly> and any question of ambiguity further than that it completely outside the realm of a simple uniform identifier
[20:35] <mikekelly> s/it/s
[20:35] <mikekelly> is
[20:35] <ordnungswidrig> mikekelly: I'm not sure if the context is sufficient. you can mix contexts
[20:35] <mikekelly> that's a feature
[20:35] <mikekelly> :)
[20:35] <mamund> context is not definitive, but it is a contributor.
[20:35] <ordnungswidrig> mikekelly: yes, but then you're back in ambiguities
[20:36] <mikekelly> when I say context I don't mean "the context of one URI"
[20:36] <mikekelly> I mean an entire #application state'
[20:36] <ordnungswidrig> hmm
[20:36] <mamund> and, let's be honest here, this whole schmeer is "time-bound" anyway...
[20:36] <mamund> what is "true" today for some URI might not be true tomorrow
[20:37] <mamund> Roy touches on this in his diss, but doesn't really get into it much.
[20:37] <mikekelly> membership function ?
[20:38] <mamund> that's one of the reasons i'm intrigued by the memento proposal
[20:38] <mamund> yeah
[20:38] <mikekelly> yeah it's just annoying we're getting yet more URIs
[20:38] <mikekelly> I asked them about how they would handle a situation where you have multiple time-gates over different periods
[20:39] <mamund> well, memento reduces the number of URIs needed to solve the history problem.
[20:39] <mikekelly> not really, people will still want unique URIs for each snapshot
[20:39] <mikekelly> all we're doing is getting more 'timegate' URIs
[20:39] <mamund> people always want stuff
[20:40] <mikekelly> well they'll, presumably, just reuse all the URIs that exist already
[20:40] <mikekelly> but I would want to do that myself anyway
[20:40] <mikekelly> for the simple reaosn that most people think a bookmark is a URI
[20:40] <mikekelly> woops bookmarking conneg'd resources breaks.
[20:40] <mikekelly> representations*
[20:40] <mikekelly> :)
[20:40] <mikekelly> I tried to raise this ages ago but I don't think I expressedd it very well
[20:42] <mikekelly> what you really should be doing when you bookmark is caching the request (or at least the parts that Vary are telling you matter)
[20:42] <mikekelly> but nobody does that
[20:43] <mamund> yep, that's an agent programming issue. we still have a long way to go on creating better web agents (of all types)
[20:43] <mikekelly> well there's lots of REST people who think that a bookmark is a URI
[20:44] <mikekelly> and back on that memento thing I don't get how the multiple timegate thing would work
[20:44] <mikekelly> if you just used conneg on the original URI it would be a lot easier :P
[20:45] <mamund> yeah, i never sat down and worked out "the math" on that percieved problem.
[20:45] <mikekelly> yeah I don't buy it, personally
[20:45] <mikekelly> :P
[20:46] yatzie (~redox@bl10-140-171.dsl.telepac.pt) joined #rest.
[20:46] yatzie (redox@bl10-140-171.dsl.telepac.pt) left #rest.
[20:47] <mikekelly> is anyone good at "statistics" ?
[20:48] <ordnungswidrig> mikekelly: it greatly depends
[20:48] <mamund> i'm good at generating them, i think.
[20:48] <mikekelly> mamund: any ideas how I could come up with some numbers on what level of efficiency increase you could get from cache invalidation as opposed to expiration
[20:48] <mikekelly> ?
[20:48] <mikekelly> I know it's a lot more efficient I just don't even know how I would start coming up with something to back that up
[20:50] <ordnungswidrig> mikekelly: is it alway more efficient? There must be work done to invalidate a cache. If 90% of the resources are requested only once then the effort of invalidating the cache might be higher than the effort if a resource expires while beeing unchanged.
[20:52] <mikekelly> well the kind of invalidation I'm talking about doesn't involve anything other than "sitting in the middle" and waiting for the client to do something which yo uknow invalidates the object
[20:52] <mamund> sorry, was out of the room...
[20:53] <mamund> i have no idea. i'd ask mnot, tho. he's been down this path before.
[20:53] <mamund> steve souders might be another good resource (i've not ever had an interaction w/ him before)
[20:54] <jsled> mikekelly: I'm not sure it's statistics & just algebra, really.
[20:55] <mikekelly> I guess you'd need stats to work out the average inefficiency of expiration
[20:55] <jsled> If you have X requests that take M bytes for request/response, and some percentage P of them are instead serviced in N bytes because of invalidation &
[20:55] <jsled> Or instead of bytes, look at time.
[20:56] <ordnungswidrig> mikekelly: ok, if you're listeing to the client for put/delete/post/patch then ok
[20:56] <ordnungswidrig> mikekelly: I tought of the backing server actively invalidating the caches by notifieng a proxy
[20:56] <mikekelly> yeah there's some proposal to that
[20:57] <mikekelly> they're just a bit shit
[20:57] <mikekelly> which isn't great
[20:57] <mikekelly> one's WCIP
[20:57] <mikekelly> there's a couple more
[20:58] <mikekelly> but they don't scale that easily
[20:59] <ordnungswidrig> what if there are multiple proxies? One will not detect the invalidation because the client used another proxy
[20:59] <mikekelly> right you need to peer your cache instances so they're a layer
[21:00] <mikekelly> invalidation only works if it's a complete layer on the server side, basically
[21:00] <mikekelly> but that's fine because you have validation for the client side
[21:01] <mikekelly> clients validate etags up to the proxy cache.. which only ever bothers the origin server when a client does something invalidatory
[21:04] <ordnungswidrig> mikekelly: ok, but gossiping between proxies would be possible, right?
[21:04] <mikekelly> right
[21:04] KevBurnsJr (~kevburnsj@c-67-161-65-81.hsd1.ca.comcast.net) joined #rest.
[21:04] <mikekelly> http://www.htcp.org/ sort of does it
[21:12] <ordnungswidrig> etag is magic, browser -> proxy -> reverse-proxy -> web-application -> couchdb
[21:12] <ordnungswidrig> etag is threaded through. couchdb is only hit if necessary.
[21:14] <ordnungswidrig> is there a best practices document for a "composed etag". I have a resource which is a collection of resources and I sha-sum the etags of the resources in the collection to get the etag of the collection.
[21:15] <mikekelly> an etag is just a string of some kind
[21:15] <mikekelly> doesn't make any difference
[21:16] <ordnungswidrig> mikekelly: I know, I meant the semantic: is it correct? If any resource in the collection changes then it's etag changed and the etag of the collection changes.
[21:17] <mikekelly> well yeah the way you're generating the etag is tied to the state of the other resources its state depends on
[21:17] <mikekelly> so you're golden
[21:18] <ordnungswidrig> fine. however if the collection resource is only a list of uris then it's questionable if the state of the collection changes.
[21:18] <ordnungswidrig> if the collection contains parts of the individual resource, e.g. title, then its clear.
[21:19] <mikekelly> right - that's called Doing It Right
[21:20] <mikekelly> except that means your clients will have to make subsequent requests for each of the items
[21:20] <mikekelly> which most people think is 'too chatty'
[21:21] <mikekelly> btw if you include something like title
[21:21] <mikekelly> technically you're still including some dependenceis
[21:23] <ordnungswidrig> yes, it's especially chatty if I can expect only a small fraction of the resources to be request after the collection.
[21:23] <ordnungswidrig> If I expect that every resource will be requested than it's more about parallel fetch, keep-alive and pipelining
[21:24] <ordnungswidrig> for a webshop product category, e.g, I'd expect every contained product to be loaded because title, picture, price etc. will be displayed for the category
[21:27] <mamund> "The Fate of the Semantic Web" http://www.pewinternet.org/Reports/2010/Semantic-Web.aspx
[21:27] <mikekelly> yeah I saw that lol :)
[21:27] <ordnungswidrig> hehe, just opened it from your twitter stream
[21:29] <mamund> not very "meaty" but it's likely to attract some attention on the press
[21:29] <mamund> from the "Methodology" section: "Since the data are based on a non.random sample, a margin of error cannot be
[21:29] <mamund> computed, and the results are not projectable to any population other than the experts in this
[21:29] <mamund> sample.Since the data are based on a non.random sample, a margin of error cannot be
[21:29] <mamund> computed, and the results are not projectable to any population other than the experts in this
[21:30] mamund dangit! double-posted _again_!
[21:30] <ordnungswidrig> mamund: post some anti-text!
[21:32] <mamund> hehehehe
[21:33] <mamund> ".elpmas siht ni strepxe eht naht rehto noitalupop yna ot elbatcejorp ton era stluser eht dna ,detupmoc
[21:33] <mamund> eb tonnac rorre fo nigram a ,elpmas modnar.non a no desab era atad eht ecniS"
[21:34] mamund was goaded into doing that by ordnungswidrig
[21:37] <ordnungswidrig> it works!
[22:30] avdi (~avdi@c-174-55-88-101.hsd1.pa.comcast.net) joined #rest.
[22:48] darrelmiller (~darrelmil@bas4-montreal45-1242355774.dsl.bell.ca) joined #rest.
[22:53] algermissen (~algermiss@p54907F83.dip.t-dialin.net) left irc: Quit: RESTing
[23:06] avdi (~avdi@c-174-55-88-101.hsd1.pa.comcast.net) left irc: Quit: avdi
[23:19] bigbluehat (~byoung@adsl-177-117-159.gsp.bellsouth.net) left irc: Quit: bigbluehat
[00:00] --- Wed May 5 2010