Collected Documentation - 2008/03/12 23:37It's tough to uncover all the functionality of Palabre, since Célio spends more time coding than documenting (not a bad thing!).
However, it is driving me crazy...};^)
So, I pulled together everything I could find that seemed relevant from:
Re:Collected Documentation - 2008/03/13 11:08Hey, congratulations ! And thanks a lot for your work !
I thought of creating a wiki for some times ... but this seems very nice. I won't have any time for the next two weeks, but as soon as I come bak I'll try to release a stable version of palabre 0.6 and then get some doc and plugins :)
keep up the good work !Célio Conort Lonesome Palabre developper
Re:Collected Documentation - 2008/03/13 16:55I apologize for the double posting, but I also wanted to say that the gathering of all of that useful data is really nice, chromecow! Thanks!
Re:Collected Documentation - 2008/03/31 10:18Hi, Thx a lot for the diagrams !
for the questions :
* How are broadcast messages originated? They are not handled by parseData.
They are handled by clientHandleMessage() But i've seen a problem, I haven't modified them since I changed the behavior of the nodes (any node can be sent as soon as they have a toroom or toclient attribute) so for now only <m> and <msg> nodes supports broadcasting. I'll change it for 0.6 stable
The doAction method seems to be acting on actions that are registered from some unknown place, like 'onClientJoinRoom'
The doAction method is called everytime an "action" is performed, and there are calls to this function from everywhere in the code :) ex :
But I think I should add some calls like : clientCanJoinRoom , ... and usse the returns of this methods.
This will make plugins like password protected rooms easier.
I'll turn it in my head (does this mean anything in english ?) and add thoose few calls and soon release 0.6 stable :)
Thanks a lot for your help and feedback !Célio Conort Lonesome Palabre developper
But I think I should add some calls like : clientCanJoinRoom , ... and usse the returns of this methods. This will make plugins like password protected rooms easier.
I'll turn it in my head (does this mean anything in english ?) and add thoose few calls and soon release 0.6 stable :)
Thanks a lot for your help and feedback !
Admin, j'aurais une suggestion. Tu pourrais regarder comment fonctionne le framework django de ce coté la.
Chercher trois fonctions dans les plug-in comme beforeAction(args)|beforeNode(args) doAction(args)|doNode(args) afterAction(args)|afterNode(args) Les exécuter si elles existent.
Ou simplement deux, un pré-traitement (tout les plug-in enregistrés pour cette action) puis le traitement de l'action.
De cette manière tu permet aux plug-in d'intervenir avant une action spécifique comme vérifier le mdp pour une chambre et la soit ne rien faire si le mot de pass est correcte, soit envoyez un message d'erreur au client et renvoyer false ce qui interromprait le processus.
Je ne sais pas si c'est possible dans ton architecture actuelle. Mais cela offrirait un système très dynamique dans ses possibilités de modifications.
Sinon dans la même optique permettre aux plug-in de s'enregistrer pour parser tout les messages sortants ( je ne sais pas si c'est possible ) de cette manière un plug-in de censure serait possible, ou un plug-in de mise en forme.
Re:Collected Documentation - 2008/04/01 17:44Please excuse the machine translation of you post, but I'm working with some of these issues currently.
I haven't found a solution for the AfterAction, but before action is pretty simple to work around.
For instance, I want to do some formating and searching of text strings before a message is sent on it way to the intended user.
So I am creating a MessageBeforeAction module, and registering it to look for the node <msgBA> instead of the standard <msg>.
The client is configured only to send these <msgBA> nodes. My module looks at them, preforms the appropriate actions, and then sends them on as standard messages.
That's the theory, anyway.
The one thing that would be super-useful is the After Action idea.
For instance, I am maintaining a list of available rooms. Currently, every client has to send a request to the server every 2 seconds to ask for an updated room list, and most of the time, they get back the list they already have.
It would be great to have an after action hook on room creation, so that when a room is created, the server could push the room to the clients, instead of the constant polling.
Admin, I have a suggestion. You could watch how the framework of this django the side.
Searching three functions in the plug-in as BeforeAction (args) | beforeNode (args) DoAction (args) | doNode (args) AfterAction (args) | afterNode (args) The execution if they exist.
Or just two, a pre-treatment (all plug-in made for this action), then the processing of the action.
In this way allows you to plug-in to intervene before a specific action as mdp check for a room and either do nothing if the password is correct, or send an error message to the client and the return false which interrupt the process.
I do not know if this is possible in your current architecture. But it would provide a very dynamic system in its ability to change.
Otherwise in the same vein allow the plug-in to register for parsing all outgoing mail (I do not know if that is possible) In this way a plug-in "would be possible, or a plug-in formatting.
Yes at first I wanted to do something like before/after/do ... But i thought it might slow down the whole server execution.
Anyway I can add many calls before/after/... very easily, it's just a server.doAction('beforeBla') bla() server.doAction('afterBla') (if you see places where thoose calls are missing just let me know)
The only problem is "what do I expect as return" ? true/false ? an edited node ? what if multiple plugins defines the same calls and are returning different results ?
Maybe I should add calls like :
if server.doAction('canDo_BlaBla',params): params = server.doAction('before_BlaBla',params) BlaBla(params) server.doAction('after_BlaBla',params)
Checking if we can, before do action, sending params and getting them possibly edited, and after action call ...
I think about that this night, and there's my solution. (mabye bad :p)
In client/room/server __init__() register function that you wanna open to plug-in use.
next create a function def __call__(self, name, params): (not sur of python syntaxe for this one) that will test if function called is registered as usable for plug-in if no -> call this function if yes -> foreach plug-in registered for this function -> params = plug-in.beforeAction(name, params) --> if params== false : return params (will stop this call)
call this function with new params.
foreach plug-in registered for this function -> params = plug-in.afterAction(name, params) --> if paramss = false : return params
And for nodes : when you know the node name and call function for : do something like this foreach plug-in registered for this node -> node = plug-in.beforeNode(node) --> if node == false -> return false (will stop the process)
standard use of this node
foreach plug-in registered for this node -> node = plug-in.AfterNode(node) -- > if node == false return false (break...)
Plug-in must extend a class named BaseModule or BasePlugin like this : def baseModule def beforeAction(name, params) return params def afterAction(name, params) return params def beforeNode(node) return node def afterNode(node) return node
like that, plug-in may rewrite some or all of this function, and that will work.
A system like that will offer a lot of possibilities for example for a game with player movement :
the client send <dep x=1 y=3 /> the plug-in checkDep will use beforeNode and check if movement is regular, if player can go there, etc.. and adds a little information to the node like <dep client=Cyprien x=1 y=3 fromx=0 fromy=2 blablabla /> next another plug-in will if this client is "a potential hacker..." log this information with beforeNode
next the normal process with this node will do nothing it's not a regular node. But next the broadcast plug-in will brodcast this node to all player in sight of fromx and fromy...
it's an example, not the best, we can do all this mess on only one plugin.
But you can add a spell checker, a "censure maker" (don't know in english), and some format plug-in with beforeNode, some checkplugin in beforeAction and beforeNode, some logger, or broadcaster in afterAction and afterNode, etc... That allow a lot of thing.
And i don't think it will slow a lot the server, because each call will just go into __call__ like a normal call and next, you will not have a lot of plugin registered, maybe like 4-5 for one node will be huge i think.
What do you think about that ?
And sorry for my poor english, i hope you will understand my words.
Re:Collected Documentation - 2008/04/06 01:25I like it. I had a hacky work around for the bad client, but it was ugly and difficult to implement, so I would love to see a system like this.
I also concur, the load on the server with this approach should be as low as the current version until people start adding their own actions.
Re:Collected Documentation - 2008/04/08 20:02While we're talking about security, there is a big hole there that needs to be filled for anything more than a hobby site, and that's the ability to create a SSL connection between the client and the server.
Otherwise, we're pretty much sending our passwords in plain text.
For the moment, I can live with this, but soon I'm going to have to tackle some of these security issues.
Has anyone thought about what it would take to make an SSL layer?
and this securised page wills end you something like UNIQUE_IDENT_ID you will send this ident to palabre for identification like connect login="toto" pass="unique-ident-id"
palabre will check in ident_table if login=xx and unique_ident_id=yy
and then (why not) delete this record from database. Like that, if someone listen your palabre conversation, he can't log-in with your unique-ident-id because it can be used only one time.
You will have secure authentification, and next a plain conversation, without using ressource for crypting/decrypt, etc...
If flash can't load https page with load_var you can put your application into a https page, that will show after a traditionnal login page (php/python (django)/etc...) with flash_application and a variable that containt login, room_to_connect (why not) and unique_ident_id
I was doing things like that for my project. Authentification in php, and next an ID that link to this authentification in database.
DOWNLOAD
FORUM
FAQ
WHAT IS IT
