Changes from Version 1 of CollabInventati

Show
Ignore:
Author:
ryan (IP: 76.210.162.198)
Timestamp:
07/20/08 11:40:50 (1 year ago)
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • CollabInventati

    v0 v1  
     1= Collaboration with Autistici-Inventati Resistant Network = 
     2 
     3Attached to this page is the A/I "Orange Book" which describes how they have set up their resistant network. I chatted with a couple of them and I think we could benefit from their CFengine ideas (and others) and the could benefit from our use of ICE. An English version of the Orange Book is attached and a transcript of our chat together follows. 
     4{{{ 
     509:14 <&bomboclat> porcodio 
     609:15 <&lesion> eh quante ne vuoi 
     709:15 < Avvelenato> perch`e usare skype??? 
     809:15 <&bomboclat> nel 2008 voglio delle interfaccie grafiche 
     909:15 < Avvelenato> non puoi usare quei grazioni programmi opensource? 
     1009:15 <&bomboclat> Avvelenato, per provare la webcam? 
     1109:15 <&bomboclat> FUORI I NOMI 
     1209:15 <&bomboclat> a parte ekiga 
     1309:15 < Avvelenato> :p 
     1409:15 <&bomboclat> io voglio streammare su icecast 2 
     1509:15 <&bomboclat> la mia faccia di cazzo 
     1609:15 < Avvelenato> io non uso skipe.. ma sono molti i programmi che lo emulano 
     1709:15 <&bomboclat> che programma uso GPL ? 
     1809:15 < Avvelenato> si 
     1909:30 -!- bomboclat [bomboclat@mufhd0-95E6567E.fastres.net] has quit [Leaving] 
     2009:35 < Avvelenato> Porco dio cane onnipotente 
     2109:36 < Avvelenato> SApete per caso se sono io ad avere problemi con il .zip  o `e noblogs? 
     2209:54 -!- babau [rizomo@mufhd0-EB72053.t2.dsl.vodafone.it] has quit [Input/output error] 
     2310:02 -!- Avvelenato [dario@mufhd0-A6E2E5B1.cust-adsl.tiscali.it] has quit [Ping timeout] 
     2410:13 < ryan> hello hello 
     2510:13 <&lesion> hello 
     2610:15 < ryan> buscando mais info di Plan R* 
     2710:15 <&lesion> que info ? 
     2810:16 < ryan> specifiches 
     2910:17 < ryan> stiamo lavorando con un progetto simile 
     3010:17 < ryan> -> http://dev.bunke.indymedia.org/ 
     3110:18 < ryan> http://linefeed.org/ 
     3210:19 <&lesion> :D 
     3310:20 <&lesion> so what kind of info do you need ? 
     3410:20 < ryan> posso falar em portugues melhor que italiano :) 
     3510:20 <&lesion> eh porugues is a problem. what about english ? 
     3610:20 < ryan> yes, english is fine 
     3710:20 <&lesion> nice 
     3810:20 < ryan> i'm interested in architecture details 
     3910:21 <&lesion> have you seen the pdf ? 
     4010:21 < ryan> no! 
     4110:21 <&lesion> the orange book I mean 
     4210:21 < ryan> i dont think so! 
     4310:24 <&lesion> ok wait 
     4410:25 < nah> http://dev.autistici.org/orangebook/ :x 
     4510:25 <&lesion> http://dev.autistici.org/orangebook/index.html.en 
     4610:25 <&lesion> yep 
     4710:25 < nah> :) 
     4810:25 <&lesion> thanks :P 
     4910:26 < nah> disponha 
     5010:27 < ryan> mm thanks 
     5110:28 < ryan> this is interesting.. 
     5210:28 < ryan> our plan involves the use of ICE 
     5310:28 < ryan> http://www.zeroc.com/ice.html 
     5410:32 < ryan> i think 8.1.3 is mostly where we are focused 
     5510:33 < ryan> because we want to build a network where all IMC sites are hosted on all the servers in the network 
     5610:33 < ryan> interesting 
     5710:34 < ryan> not exactly what we have planned but -we- can definitely use parts of what you have (especially the cfengine stuff) 
     5810:34 < ryan> and this orange book could contain some of the parts that we have (especially the ICE stuff) 
     5910:34 < ryan> is this being worked on? 
     6010:35 < ryan> is this network implemented in any way? 
     6110:35 <&megabug> autistici/inventati is a working example 
     6210:36 <&megabug> lesion ! tesoro :) 
     6310:36 < ryan> how many servers are in the network? 
     6410:36 <&megabug> 9 actually 
     6510:37 <&megabug> 4 are the main server for services like web mail etc. 
     6610:37 <&megabug> and 5 servers are for backup / monitor / statistics and minor services 
     6710:37 <&lesion> 5 
     6810:37 <&megabug> ah! right 5 :P 
     6910:38 < ryan> are any of them supporting italy indymedia? 
     7010:38 <&lesion> nope 
     7110:38 <&lesion> to host a web service replicated on a lot of hosts the key point is replicate database stuff 
     7210:38 < ryan> our main objective is to build a network such that if we suffer server seizure (by law enforcement), hardware failure or donation loss (like ahimsa), 
     7310:39 < ryan> there is no interruption in service 
     7410:39 <&lesion> cool 
     7510:39 < ryan> yes, we manage the data with ICE 
     7610:39 < ryan> so ICE is an object-store for us 
     7710:39 < ryan> using a database as a backup 
     7810:40 <&lesion> but have you developed the indymedia cms to use ice ? 
     7910:40 < ryan> well, this is what we're working on 
     8010:40 < ryan> we decided to merge the indymedia CMS teams 
     8110:40 <&lesion> cool 
     8210:40 < ryan> at least, mir & sf-active 
     8310:40 < ryan> because we were stretched too thin 
     8410:40 <&lesion> nice idea :D 
     8510:41 <&megabug> argh :) 
     8610:41 < ryan> so we are working on a new front end, which i believe will be something like joomla, extended by cakephp 
     8710:41 <&lesion> noooooooooo joomla buuuuuuuuu 
     8810:41 <&lesion> :) 
     8910:41 < ryan> why not? 
     9010:42 < ryan> the days of mir and sf-active are numbered 
     9110:42 < ryan> soon, they will be only memories 
     9210:42 <&megabug> how much time take a new front end to be developed? 
     9310:43 < ryan> well, if we begin with an existing one, not much time at all :) 
     9410:43 < ryan> our point was to find a CMS which already has a large development community 
     9510:43 < ryan> but, none of them were very good 
     9610:43 < ryan> so, then we thought we would write our own, using CakePHP 
     9710:44 < ryan> but, we didnt want to write from scratch 
     9810:44 < ryan> then, we found a Joomla<->CakePHP bridge  
     9910:44 < ryan> and i think we will use that  
     10010:44 < ryan> so we can get all the benefits of the large development community of joomla AND cakephp 
     10110:45 <&lesion> adapting joomla to your needs will need more time than develop a new front end from scratch imho 
     10210:45 < ryan> how so? 
     10310:45 <&lesion> eh you will see ;) 
     10410:45 < ryan> nah 
     10510:46 <&lesion> hope so 
     10610:46 <&megabug> uhm i don't now, how much joomla is good for you? 
     10710:46 < ryan> who will make the admin interface for our cms-from-scratch? 
     10810:46 < ryan> design the nice interface? 
     10910:46 < ryan> make all the extensions? 
     11010:46 < ryan> etc etc 
     11110:46 < ryan> i've written an indymedia cms from scratch before, so i know how it goes 
     11210:46 < ryan> for instance -> http://extensions.joomla.org/ 
     11310:46 <&lesion> ok but now you have experience with you :D 
     11410:46 <&megabug> maybe you can write plugins 
     11510:47 < ryan> that would -never- happen with our own cms 
     11610:47 < ryan> yes, and my experience tells me: reusable code is king :) 
     11710:47 <&megabug> to giva joomla the features needed by imc 
     11810:48 <&megabug> give 
     11910:48 <&lesion> but also, modify others code is a pain  
     12010:48 < ryan> but, what features an imc needs that joomla does not have? 
     12110:49 <&megabug> i don't know :) 
     12210:49 < ryan> none! 
     12310:49 < ryan> in fact, it offers much more than we could build ourselves 
     12410:49 <&megabug> you tell us that you want to developo with cakephp :) 
     12510:49 <&megabug> and joomla 
     12610:49 < ryan> we want to extend with cakephp 
     12710:50 < ryan> to develop the ICE interface 
     12810:50 <&lesion> ah diobono vero 
     12910:50 < ryan> and, if we want to add new features, we can do it 
     13010:50 <&megabug> ah ok 
     13110:50 < ryan> for instance, joomla currently has 27 extensions for listing articles 
     13210:50 <&megabug> but why ice? 
     13310:51 <&lesion> to spread contents over servers 
     13410:51 < ryan> i havent found something else as good as ice 
     13510:51 <&lesion> mysql replication ?  
     13610:52 < ryan> we will also use mysql replication 
     13710:52 < ryan> but ice will act between the web server tier and the db tier 
     13810:52 < ryan> for instance, mysql is a poor and slow object store 
     13910:52 < ryan> ice is a fast object store 
     14010:54 <&megabug> you want to switch from mysql to ice? 
     14110:54 <&megabug> i don't know how ice does really 
     14210:54 < ryan> ice is middleware 
     14310:55 < ryan> it sits between the web tier and the db tier 
     14410:55 < ryan> so, for instance, if a user logs into Web Server #10 
     14510:55 < ryan> it stores the session, via ice, in mysql but ice remembers it 
     14610:55 < ryan> so, when the user returns to Web Server #4, and asks for the session, 
     14710:55 <&megabug> ok 
     14810:55 < ryan> ice gives it the session object 
     14910:56 < ryan> without going to mysql 
     15010:56 <&megabug> ah interesting 
     15110:56 < ryan> ice also understands where all the webservers are 
     15210:56 <&megabug> but in your replication model 
     15310:56 < ryan> and where all the dbservers are 
     15410:57 <&megabug> replicating sessions is a different level from replicating data 
     15510:57 < ryan> what do you mean? 
     15610:57 <&megabug> sessions are not stored on mysql 
     15710:58 < ryan> we intend to 
     15810:58 <&megabug> is a web server issue 
     15910:58 <&megabug> we got the same session problem with our services 
     16010:59 < ryan> i read about it in the orange book 
     16110:59 < ryan> you use sticky sessions 
     16210:59 < ryan> but if you store the sessions in the db, you dont have this problem 
     16310:59 <&megabug> yes maybe we can reconsider the solution 
     16410:59 < ryan> yes, sticky sessions mess up your load balancing 
     16511:00 <&megabug> mmmhh, not really 
     16611:00 <&megabug> with an high number of users the difference is not significant 
     16711:00 < ryan> well, the point of round-robin dns is so that hits to the server can be all over the place, right? 
     16811:01 <&megabug> yep 
     16911:01 < ryan> but if you have a number of heavy users "stuck" to a particular web server, then you have defeated this purpose 
     17011:02 < ryan> by "heavy users" i mean... 4-5 users uploading 50MB videos to an imc all at once 
     17111:02 <&megabug> yep but its the worst case scenario 
     17211:02 <&megabug> its very unlike to happen 
     17311:02 < ryan> i think it would be quite common with IMC, especially during a period of heavy usage (like a big protest) 
     17411:02 < ryan> how unlikely it is to happen depends on how many webservers you have :) 
     17511:02 <&megabug> and users... 
     17611:02 < ryan> if it is 2, it is very likely to happen 
     17711:03 <&megabug> uhm 
     17811:03 <&megabug> it's 1 possibility on 2^users 
     17911:03 < ryan> well.. our goal is to host over 140 websites on the network 
     18011:03 <&megabug> (when all the 5 uploaders match the same server) 
     18111:03 < ryan> all of them receiving or sending large media files 
     18211:04 < ryan> so we want to make sure that we get the advantage of round-robin dns 
     18311:04 <&megabug> the balancing with rrdns is very good 
     18411:04 <&megabug> the problem is another 
     18511:05 <&megabug> and is that not all the user make the same use of services 
     18611:05 <&megabug> so you got 500 lurker and 1 heavy uploader for example 
     18711:06 <&megabug> in this case this single uploader can exausth all resources on a server 
     18811:07 <&megabug> but it's very rare, we got similar problem 6/7 times in 5 years 
     18911:08 < ryan> sure, that could happen 
     19011:08 < ryan> but it's even more likely to happen with the model we have now which is 15 IMC's on one webserver :) 
     19111:08 <&megabug> ghghghgh 
     19211:08 <&megabug> yes i imagine :D 
     19311:09 < ryan> well, this is interesting.. i think we will definitely want to use your cfengine stuff 
     19411:10 < ryan> and maybe you will be interested in how we use ICE for web applications 
     19511:10 <&lesion> yep 
     19611:10 <&lesion> we have to replicate lifetype 
     19711:10 <&megabug> yep, for our front-end maybe interesting 
     19811:10 < ryan> who are the main people who work on this for a/i? 
     19911:10 <&megabug> for user websites maybe not :) but we'll see 
     20011:10 <&lesion> on cfengine stuff?  
     20111:11 < ryan> well, yeah, just the whole idea of the resistant network? 
     20211:11 -!- lesion [lesion@mufhd0-6BDA3380.ip.fastwebnet.it] has quit [.] 
     20311:11 <&megabug> the r* come from A/I collective altogheter 
     20411:11 < ryan> also do you mind if i share this chat with other people in our group working on this? 
     20511:12 <&megabug> from discussion internal to the group 
     20611:12 <&megabug> yes of course 
     20711:12 <&megabug> all R* stuff is public :) 
     20811:12 < ryan> nice, same with our stuff 
     20911:12 < ryan> also we are meeting in rio de janeiro  
     21011:12 < ryan> in september 
     21111:13 < ryan> to advance this project 
     21211:13 < ryan> at techmeet 
     21311:13 < ryan> (techmeet.org) 
     21411:13 < ryan> the groups involved in this project are: 
     21511:13 <&megabug> cool :) 
     21611:13 < ryan> techmeet.org 
     21711:13 < ryan> linefeed.org 
     21811:13 < ryan> imc-cms group in indymedia, lists.indymedia.org/imc-cms 
     21911:13 < ryan> dev.bunke.indymedia.org 
     22011:14 < ryan> come to rio! 
     22111:14 < ryan> :) 
     222 
     223}}} 
     224