User:Which Linden/Office Hours/2007 Sep 20

From Second Life Wiki
< User:Which Linden/Office Hours
Revision as of 13:42, 20 September 2007 by Which Linden (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Transcript of Which Linden's office hours:

[11:12] Which Linden: Incidentally, I plan to post this transcript if it's cool with you all
[11:12] Tao Takashi: sure :)
[11:12] Timeless Prototype: fine with me
[11:12] Which Linden: OK, that'll be the rule for future office hours
[11:12] Tao Takashi: so I will hopefully find out more about this egg thing ;-)
[11:13] Laetizia Coronet is not sure what it is about anyway ;)
[11:13] Which Linden: The egg thing is totally on my task list
[11:13] Tao Takashi: but I think developing and having it installed on the same machine should be no problem
[11:13] Tao Takashi: PYTHONPATH should still be evaluated
[11:13] Which Linden: Laetizia: python eggs are basically a convenient way to install Python code
[11:13] Tao Takashi: and as said for developing buildout might be of great use
[11:14] Which Linden: Kind of like Perl's CPAN and Ruby Gems
[11:14] Tao Takashi: just give somebody a buildout script and it automatically installs all components needed for running the project
[11:14] Which Linden: Hm.... that is nice
[11:14] Laetizia Coronet: ok... then this is no office hour for me - I was witing for Daniel...
[11:14] Tao Takashi: be it from svn, rpm, tgz, http, ftp whatever
[11:14] Which Linden: Laetizia: oh, no Daniel office hours
[11:14] Which Linden: We'd need to have eggs for Eventlet and Mulib, too
[11:15] Laetizia Coronet: no idea what Python is beyond Monty :) have a good evening
[11:15] Which Linden: You too
[11:15] Tao Takashi: you too, Laetizia :)
[11:15] Kitto Flora: Did I find the supposed Daniel Meeting?
[11:15] Tao Takashi: Python is the best thing on earth though
[11:15] Which Linden: Hah, wow, no, no Daniel here
[11:15] Laetizia Coronet: No Daniel on his spot
[11:15] Laetizia Coronet: Levenhall, the cornfield. Noone there
[11:15] Tao Takashi: Daniel should be somewhere else but I also waited once for him without him coming
[11:15] Tao Takashi: like with Glenn earlier
[11:15] Kitto Flora: Where Daniel meets is a secret?
[11:16] Tao Takashi: and for choosing the storage, zope3 interfaces might be nice
[11:16] Laetizia Coronet: no no it is in Levenhall but e has a bad record for not showing up
[11:16] Laetizia Coronet: bye
[11:16] Which Linden: see ya!
[11:17] Tao Takashi: so you basically declare an interface like IPersistenceManager and then you implement your persister class with implements(IPersistenceManager)
[11:17] Tao Takashi: and then you can register it as a utility and retrieve it again with getUtility(IPersistenceManager)
[11:17] Which Linden: What this reminded me of was that we still are working towards 1.0 with Eventlet and Mulib
[11:17] Tao Takashi: which returns the actual configured implementation of IPersistenceManager
[11:17] Tao Takashi: this has a set of defined methods then
[11:17] Which Linden: Interesting, I'll give it a look
[11:18] Tao Takashi: we really should do Python user Group meetings in here
[11:18] Tao Takashi: I should join that group again and maybe announce something on the mailinglist
[11:18] Tao Takashi: another task I probably won't have time for thouggh ;-)
[11:18] Tao Takashi: in Zope3 we also have the nice concept of adapters
[11:19] Tao Takashi: like if you have a class which does not implement a specific interface, like a getSize() method of an ISize interface
[11:19] Which Linden: So check this page out: http://wiki.secondlife.com/wiki/Eventlet_1.0
[11:19] Tao Takashi: thn you can write an adapter which adapts the instance of this class to this interface
[11:19] Tao Takashi: simply by writing adapater=ISize(instance)
[11:19] Tao Takashi: and then adapter.getSize()
[11:20] Tao Takashi: which is quite handy and better than subclassing
[11:20] Tao Takashi: because you can configure the correct adapter and you don't need to change the original code of the class
[11:20] Which Linden: Also: http://wiki.secondlife.com/wiki/Mulib_1.0
[11:20] Tao Takashi: *checking* :)
[11:21] Which Linden: They're just a bunch of links to JIRA tasks, but basically what we're asking for is additions
[11:21] Which Linden: I know people haven't really gotten off the ground with these libraries
[11:21] Tao Takashi: I think they are quite specific for a certain task
[11:21] Which Linden: (except for Baba who's tackling the hardest problem head-on)
[11:22] Which Linden: Hmmmm.... I tend to think of them as quite general, but maybe that's because I have lots of use cases. :-)
[11:22] Which Linden: What's the task you see them as limited to?
[11:22] Tao Takashi: I must admit that I only skimmed them by now ;-) maybe they also simply need more PR
[11:22] Tao Takashi: I need more time ;-)
[11:22] Which Linden: Definitely need that
[11:22] Which Linden: I need more sleep, while we're wishing. :-)
[11:23] Timeless Prototype needs more brains, hands and computers
[11:23] Which Linden: So what should we do for a start? Expand the examples more?
[11:23] Tao Takashi: looking at the examples now :)
[11:23] Which Linden: hah
[11:23] Tao Takashi: do you do doctests?
[11:23] Which Linden: Or maybe include the examples on the front wiki page
[11:24] Timeless Prototype: the examples are bringing clarity to the intent, more examples would be cool
[11:24] Tao Takashi: yes, usually code says more ;-)
[11:24] Tao Takashi: that's also why I like doctests more because these are basically documented examples
[11:24] Which Linden: Timeless: Gotta be careful, though: too many computers is like too many cats: always one getting out of the bag
[11:24] Which Linden: What are doctests?
[11:24] Timeless Prototype: :)
[11:24] Which Linden: *googles*
[11:25] Which Linden: oh, rad!
[11:25] Which Linden: I see, it looks through your source for interactive examples then tries to run them
[11:25] Which Linden: Yeah, we should totally do that
[11:25] Tao Takashi: you can even use them in docstrings
[11:26] Tao Takashi: Plone is mainly doctests these days, explaining interfaces and APIs
[11:26] Which Linden: One problem is that these are networking libraries, so not everything makes sense as an interactive operation
[11:26] Which Linden: I.e. our examples on the wiki use curl
[11:27] Which Linden: I guess you can shell out
[11:27] Which Linden: os.system
[11:27] Tao Takashi: ok, right, there might be limits
[11:27] Timeless Prototype: btw, loved the curl examples, very concise
[11:27] Tao Takashi: and it seems always the biggest problem to work around these limits in tests ;-)
[11:28] Which Linden: Donovan came up with the examples
[11:28] Which Linden: That dude's tagline should be "it's *incredibly* simple"
[11:28] Timeless Prototype: :)
[11:28] Which Linden: We do have unit tests that start up httpd servers for each test
[11:29] Which Linden makes a task to convert to/add doctests.
[11:30] Which Linden: So for chttp, there's this dude who's been helping us with it
[11:30] Tao Takashi sees the name Bob Ippolito in the source :)
[11:31] Which Linden: He should send an intro email to sldev soon.
[11:31] Which Linden: But just so you know, and aren't confused about why seeping.blister is committing to chttp :-)
[11:31] Which Linden: (oh, it's not Bob)
[11:31] Which Linden: Bob is cool, I *wish* he was working on stuff with us
[11:32] Tao Takashi: well, just checking the source ;-)
[11:32] Which Linden: Yeah
[11:32] Tao Takashi: Bob should definitely be cool, heard lots of cool stuff about Mochikit
[11:33] Which Linden: Yeah, mochikit is really nice
[11:33] Tao Takashi: and I know we had a lengthy discussion about which of these libs to ship with Plone
[11:33] Which Linden: And apparently he has something that makes Erlang easier to use, too
[11:33] Tao Takashi never had the need to use Erlang
[11:34] Which Linden: You don't need to use Erlang, Erlang needs to use *you*!
[11:34] Which Linden: (in Soviet Russia)
[11:34] Tao Takashi: ah, ic :-)
[11:34] Which Linden: Timeless, you've been pretty quiet, anything on your mind?
[11:35] Timeless Prototype: IMs!
[11:35] Timeless Prototype: lol
[11:35] Which Linden: Hah
[11:35] Tao Takashi: btw, for the pages you might wan to explain maybe in how these libraries are different from the existing ones
[11:35] Which Linden: Yeah, well, what are the 'competitors'?
[11:35] Which Linden: Twisted
[11:36] Which Linden: httplib
[11:36] Tao Takashi: asyncore?
[11:36] Which Linden: right
[11:36] Which Linden: and mulib sort of gets into django and web frameworks territory
[11:37] Tao Takashi: yes
[11:37] Tao Takashi: at least in a basic sense
[11:37] Which Linden: The major thing about eventlet and mulib is that they make it very easy to do lots of data-passing via http
[11:38] Which Linden: Using it is rather a lot like very easy rpc
[11:39] Which Linden: We should totally put up an example of a client-server application
[11:39] Timeless Prototype: yeah
[11:39] Tao Takashi: maybe a use case, too :-)
[11:39] Timeless Prototype: the things it can hook into - have they been specced yet or are you just thrashing out the mechanism first?
[11:40] Which Linden: Also, as Donovan alluded to, we have libraries that make it possible to speak LLSD with some process, which could potentially mean an all-Python client
[11:40] Which Linden: s/client/viewer/
[11:40] Tao Takashi: Xan was telling me that during the architecture meeting
[11:40] Which Linden: What do you mean "the things it could hook into"?
[11:40] Tao Takashi: that there are LLSD libs around
[11:40] Tao Takashi: I geuss the use case is backbone
[11:40] Tao Takashi: but the page is empty yet ;-)
[11:40] Which Linden: we just have to figure out the licensing for the llsd libs
[11:41] Timeless Prototype: I've been away, not followed all this stuff about mulib etc
[11:41] Which Linden: Oh, hah, yeah, the backbone page is so sad
[11:41] Tao Takashi: booking gigs for the spotlight I guess :)
[11:41] Which Linden: I don't know if backbone is really a good example, it's really just a registration service for mulib web applications
[11:42] Which Linden: so we have all these independent applications and the backbone coordinates the url space
[11:42] Tao Takashi: I wonder if I can put my PDF generator on eventlet or so
[11:42] Tao Takashi: as it really just needs a very basic web server which takes arguments and calls the PDF generation method
[11:42] Tao Takashi: right now I use TurboGears for it
[11:43] Which Linden: Sounds like a great use case
[11:43] Tao Takashi: but I don't need any mysql, templating etc.
[11:43] Which Linden: Yeah, you just want it to serve up some HTTP
[11:43] Tao Takashi: does mulib expose methods, too?
[11:43] Which Linden: Hey Armand
[11:44] Tao Takashi: or just a handle_get?
[11:44] Tao Takashi: Hi armand
[11:44] Which Linden: Oh, yeah, handle_put, handle_head
[11:44] Armand Callisto: Greetings all
[11:44] Which Linden: you name it
[11:44] Timeless Prototype: hi Armand :)
[11:44] Tao Takashi: ok, but it's not exposing methods then
[11:44] Which Linden: What do you mean "exposing" them?
[11:44] Tao Takashi: like http://localhost:8080/whatever calls whatever() in my class
[11:44] Which Linden: Oh, yeah, it does that
[11:44] Tao Takashi: ok :)
[11:44] Armand Callisto: you mean like posting the proof of concept on the jira?
[11:45] Which Linden: Armand: we're talking about using mulib for web applications, and how best to use it
[11:45] Armand Callisto: oh sweet
[11:45] Tao Takashi: Hey Donova :)
[11:45] Tao Takashi: n
[11:45] Tao Takashi: sorry :)
[11:45] Timeless Prototype: hi Donovan!
[11:45] Donovan Linden: howdy
[11:45] Which Linden: Yo dude
[11:46] Tao Takashi: so now that I know your lastname, it rings a bell :-)
[11:46] Timeless Prototype laughs
[11:46] Which Linden: Tao was just asking about how to get http://localhost:8080/whatever calls whatever() in my class
[11:46] Which Linden: ok, I clearly just copied and pasted that line, badly
[11:46] Which Linden is like eliza
[11:47] Tao Takashi: well, I could actually simply do it myself in handle_get() via getattr()
[11:47] Which Linden: So, I think you're right, you don't expose methods that way, you expose resources
[11:47] Which Linden: Because the url is a noun
[11:48] Which Linden: So you'll want to do different things whether you GET localhost/whatever or PUT
[11:48] Which Linden: or whatever the verb is
[11:48] Which Linden: That is The REST Way (tm)
[11:48] Tao Takashi: the method is a noun
[11:48] Donovan Linden: yeah -- I never liked zope putting the method name in the url
[11:48] Tao Takashi: err verb :)
[11:49] Armand Callisto: ok
[11:49] Which Linden: Yeah, the method is a verb, so each object has 6 verbs
[11:49] Which Linden: (barring http extensions like webdav)
[11:49] Tao Takashi will never come back if there are Zope haters around here ;-)
[11:50] Tao Takashi: but actually somebody implemented REST for Zope3 at the last Snow Sprint
[11:50] Which Linden: No hate here. :-0
[11:50] Tao Takashi: jk
[11:51] Which Linden: Even Rails has started to implement noun-centric REST apis
[11:51] Tao Takashi: the only thing is that I am not sure how to do it the REST way with my setup ;-)
[11:51] Tao Takashi: as this server really just get's the parameters needed for generating the PDF from the browser
[11:51] Which Linden: Yeah, it's tricky to convert existing applications
[11:51] Tao Takashi: or I construct a filename out of them but that might be long
[11:52] Which Linden: Well, you could have the client POST to get a uuid-based document url
[11:52] Tao Takashi: you mean 2 requests then?
[11:52] Which Linden: then it PUTs to that url the document, and then GETs the url with varying accept headers to choose PDF or original
[11:52] Which Linden: yea, two requests
[11:53] Donovan Linden: I can imagine a PUT of an xml or json content type, and a GET of a pdf content type doing the conversion
[11:53] Tao Takashi: well, it's really the last step in a wizard here, one server does all the wizard work in asking questions about how it should look and gets back with a big dictionary which I put in a hidden field in a form
[11:53] Donovan Linden: yeah, what which said. hehe
[11:54] Tao Takashi: and the last click will then go to another server with this information
[11:54] Tao Takashi: so with the two request I'd need some way of storing data I guess
[11:54] Donovan Linden: the first server can do the PUT to the second server, then tell the browser to GET it
[11:54] Tao Takashi: true, but wouldn't I need some temporary storage?
[11:55] Which Linden: Our development paradigm has been to build a REST backend and then build a front-end that uses HTML
[11:55] Which Linden: So, yeah, in that case the front-end has to have some state
[11:55] Which Linden: Or it could pass it around on every request
[11:55] Donovan Linden: the pdf server would be a persistent store of those xml documents, unless someone deleted them
[11:55] Which Linden: Or, well, only one request, I guess
[11:56] Tao Takashi: well, if I would generate the PDF on PUT it would work by simply refering to it
[11:56] Donovan Linden: yeah exactly
[11:56] Tao Takashi: but actually I wanted to get rid of waiting for the PDF on the original server and a PUT would do that again
[11:57] Which Linden: I guess the PDF itself could be the body of the response to the PUT
[11:57] Tao Takashi: so I guess for now I will simply stay with parameters :)
[11:57] Tao Takashi: it's not a publci API anyway here
[11:57] Tao Takashi: and also mainly just a little test
[11:58] Tao Takashi: so your use case for mulib is mainly to serve an URL dict then?
[11:58] Which Linden: Hoo, no, not really
[11:58] Which Linden: Though that ability is very useful
[11:58] Tao Takashi: see, you should write use cases :)
[11:58] Which Linden: Yeah, we should
[11:59] Timeless Prototype nods
[11:59] Donovan Linden: "REST" is the use case :)
[11:59] Which Linden: Yeah, but that's not enough
[11:59] Tao Takashi: and REST wasn't possible with other frameworks?
[11:59] Which Linden: Aren't you the one who said that Zope *just* got support for REST?
[11:59] Donovan Linden: it is possible, but most frameworks don't have it as the central design principle
[11:59] Which Linden: You have to convince people that REST is a good way to architect things
[11:59] Which Linden: :-)
[12:00] Tao Takashi: I like the concept of resting actually ;-)
[12:00] Tao Takashi: I should do that more :-)
[12:00] Armand Callisto laughs
[12:00] Donovan Linden: I think a major part of REST that people don't understand the power of is content type negotiation
[12:00] Tao Takashi: maybe I am one of them ;-)
[12:01] Tao Takashi: maybe I am too much Zope ;-)
[12:01] Donovan Linden: for example PUT as json and GET as PDF or HTML
[12:01] Which Linden: It's cause everyone is using their browser, which they can't control the Accept header for
[12:01] Donovan Linden: very powerful when you start thinking in that way
[12:01] Tao Takashi: sounds a bit like different views in Zope3
[12:02] Donovan Linden: yup
[12:02] Tao Takashi: depending on what interface is used for the incoming request (can also be ftp or whatever) content is delivered differently
[12:02] Which Linden: I think mulib really scores on negotiation. I don't even think about it
[12:02] Which Linden: (except when it's broken)
[12:02] Armand Callisto: ....nothing is broken...
[12:02] Tao Takashi: don't break it then ;-)
[12:03] Donovan Linden: Leonard Richardson gave me a good idea about content negotiation in the browser... let the file extension override
[12:03] Which Linden: Yeah, that was a good idea
[12:03] Donovan Linden: for example, PUT to foo.json, then get from foo.pdf
[12:03] Donovan Linden: the actual entity url is foo
[12:03] Which Linden: sucks that you then have to have a mapping from extension to mime-type
[12:04] Donovan Linden: python already has that
[12:04] Which Linden: Oh, well, then
[12:04] Tao Takashi: I think we need to get more Python folks here :)
[12:04] Donovan Linden: example in mulib.resources.File
[12:04] Which Linden: OK then, that's the goal: moar python
[12:05] Armand Callisto: you have a few thousand hard core python folks...just put out an invitation
[12:05] Tao Takashi: and when will you hen start with the pyhton only viewer? :)
[12:05] Donovan Linden: content_type, encoding = mimetypes.guess_type(path)
[12:05] Tao Takashi: or when is it finished is the better question
[12:05] Which Linden: I don't think we have the cycles for a python-only viewer yet
[12:05] Tao Takashi: then I would finally join the viewer development :)
[12:06] Tao Takashi: I'd have some ideas ;-)
[12:06] Which Linden: We'd just make the capability available
[12:06] Donovan Linden: viewer is probably going to stay c++ forever :)
[12:06] Tao Takashi: well, your viewer
[12:06] Donovan Linden: a python viewer would probably be a fork
[12:06] Donovan Linden: zactly
[12:06] Tao Takashi: but a library would be nice for a start
[12:06] Which Linden: True
[12:06] Which Linden: Hey, but our hour is up, so let's pick this up again next week.
[12:06] Tao Takashi: I am not really keen on doing things in C# actually ;-)
[12:06] Armand Callisto: k
[12:07] Tao Takashi: yep, I also need to go home :)
[12:07] Tao Takashi: thanks for hosting :)
[12:07] Donovan Linden: bye :)
[12:07] Timeless Prototype: k tc guys, ty
[12:07] Armand Callisto: Safe journies
[12:07] Which Linden: Thanks all, you were a gret crowd
[12:07] Tao Takashi: thanks :)
[12:07] Which Linden: great
[12:07] Which Linden: Blah, see you later
[12:07] Tao Takashi: :)
[12:07] Tao Takashi: cya