incubator-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Donald Whytock <dwhyt...@gmail.com>
Subject Re: [PROPOSAL] Chatterbot, a lightweight framework for chat responders
Date Mon, 01 Mar 2010 11:54:51 GMT
Hello again...

Following is the revised proposal text, as posted on the wiki.
Significant changes are the goals, which now focus on building the
framework around Felix and devising a standard for protocol handlers
to be used both inside and outside the project, and the committer
list, which now includes Christopher Brind.  The wiki copy is at

http://wiki.apache.org/incubator/ChatterbotProposal

Thanks...

Don

-----

Proposal:

Abstract

ChatterBot is a lightweight, multiprotocol framework and container for
chat responders.

Proposal

ChatterBot is a framework for developing chat responders (applications
that respond to messages received via online chat) and a container for
deploying them.  It is written in Java SE, and runs as a Java
application.  Chat responders are built by extending a single class
and modifying a configuration file to reference the new class.
ChatterBot's focus is on the following characteristics:

 * Small: The current framework consists of eight core classes.
 * Standalone: ChatterBot does not require external servers to operate.
 * Portable: ChatterBot should work as run from any Java-capable
machine.  For full functionality that machine should have internet
access, but localhost and console connectivity are possible.  It
should be possible to run multiple instances of ChatterBot on the same
machine or on separate machines with no loss of functionality.
 * Extensible: An instance of ChatterBot can support multiple message
parsers and protocols.  Adding more is done by editing a configuration
file.
 * Dynamic: Activating and de-activating modules should be possible
during runtime.
 * Multi-user access: Multiple users, over multiple protocols, should
be able to access deployed applications.

Rationale

A chat responder can serve as a user interface to applications, either
those built into the responder or external applications with which the
responder communicates.  Such an interface is more secure than
interfaces such as Telnet or web services since it does not require
open ports in the firewall; the container connects out through the
firewall to the chat server, rather than allowing users to connect in.

A lightweight chat responder can be installed on any system to allow
command-line access to users over whatever protocol a user may have
access to.  Thus applications can be accessed from web interfaces,
instant-message systems, text messages, email, etc.  A scalable
container can allow as many or as few access protocols as are desired.

ChatterBot, therefore, has value for those circumstances where a user
interface is needed but a server-based or enterprise solution is
either not possible or not desired.  It also can serve as a bridge
between applications, where one or more uses a chat protocol such as
XMPP to communicate.

Background

ChatterBot began in 2005 as a thin-server approach to online
multi-user board games, implemented as applets sending gamestate
changes to one another via message relaying.  The idea was to make as
general-purpose a server as possible, so that multiple games could be
built that employed the same message-relaying system.

Version 0.2 of the server was then refined in 2008 to allow for more
varied and functional message-handlers, and was used to implement a
room system that allowed for room-specific applications -- parsers
that checked the user's room before handling a command and sent
responses to other room occupants.  This version was structured
entirely around XMPP.  The global namespace was introduced to allow
modules to communicate with relatively limited coupling.

Version, 0.3, as of late 2009, functions with XMPP and has the
capacity to function with whatever other protocols channels are coded
for.  V0.3, though, uses a custom shell, with rudimentary module
lifecycle capability.

This proposal introduces version 0.4, to be based on OSGi for module
lifecycle management and event-driven module synchronization.
Applications originally built for v0.2 will be ported to v0.4.

Current Status

Meritocracy:

Peer review and alternate ideas are welcomed in this project with open
arms.  This project was intended specifically as an alternative to
traditional server-based or enterprise architecture; however, it is
recognized that tried-and-tested principles established in enterprise
architecture may be applicable here.

Core Developers:

As of late 2009, there is one developer, Donald Whytock (dwhytock at
gmail dot com).  Donald Whytock has several years of experience as a
software developer, working in a variety of languages, including C,
Java, Perl, PHP, JavaScript and SQL.  He develops both professionally
and casually; ChatterBot has been an independent, voluntary effort.

Alignment:

ChatterBot's primary potential alignment with ASF is that of a
framework for internet-accessible applications.  As command parsers
can be built to interface with other applications, ChatterBot can be
employed as a general-purpose remote console operating over instant
messages.

ASF projects we anticipate using in ChatterBot include:

 * Felix: to replace the v0.3 Shell as a module lifecycle management framework
 * UIMA: to aid in parsing/analyzing messages, either in individual
command parsers or in the parser dispatcher

Initial Goals

ChatterBot v0.3 exists as a functioning prototype, but does not
conform to existing standards in many areas, and needs expansion in
its functionality.  To this end, the project recognizes the following
goals:

 * Conversion to Apache Felix as a core framework.  This will replace
the existing Shell and affect all other modules.
 * Proposal of a standard for chat-protocol handlers, independent of
ChatterBot's specific needs.
 * Development of handlers for multiple chat protocols, compliant with
the proposed standard, for use by ChatterBot.
 * Identification/development and assembly of tools useful for
parsing, interpreting and processing text commands.

Known Risks

Orphaned Products:

Currently the project has only two committers, though Donald Whytock
has been working on the code for a few years and is committed to
seeing a functional product available.

Inexperience with Open Source:

While the developer has experience working with open-source products,
this is the first time opening up a project for open-source
collaboration.  As modular as the project is, however, open-source
collaboration should not be a problem.  It is greatly desired that
this project not be developed in a vacuum.

Fascination with the Apache Brand:

Association with the Apache brand is not sought for personal
publicity; rather, it is sought for the associated community and
access to collaboration and peer review.  This project will see its
full potential through public use and refinement, and a product more
refined for everyone's use is a more refined product for the
developer's use as well.

Initial Source

Original code developed by Donald Whytock.

Required Resources

Mailing Lists:

 * chatterbot-private
 * chatterbot-dev

Subversion Directory:

https://svn.apache.org/repos/asf/incubator/chatterbot

Issue Tracking:

JIRA ChatterBot

Initial Committers

Christopher Brind (brindy at brindy dot org dot uk)
Donald Whytock (dwhytock at gmail dot com)



On Mon, Feb 1, 2010 at 2:03 PM, Christopher Brind <brindy@brindy.org.uk> wrote:
> Sorry for shaking things up, but it sounds like you got the gist of things.
>  Using OSGi services to wire up Chatterbot makes it much more flexible in
> the long run allowing developers/users to plug in alternative
> implementations of things if they want to.  I'm quite happy to join your
> project as a committer to help guide this if you wish. :)
>
> Not sure about the proposal side of things, I'm sure someone will pipe up
> soon enough.
>
> Cheers,
> Chris
>
> On 1 February 2010 18:39, Donald Whytock <dwhytock@gmail.com> wrote:
>
>> I had originally thought that Felix Shell would replace Chatterbot
>> Listener, but I no longer think so.  Felix Shell, as far as I can
>> tell, is focused around Commands that have single outputs directed
>> toward their originator; Chatterbot Parsers, in a multiuser
>> environment, might have multiple outputs, and therefore have to
>> respond in the context of the originator. (v0.2 had writeMsg(target,
>> message) as well as writeMsgToAllBut(target, message).)  On the other
>> hand, I can see a Parser that acts like a Remote Shell.
>>
>> So at this point we're talking about changing the proposal to focus on:
>>
>> - Building Chatterbot around Felix as a modularity framework, with its
>> lifecycle management, its ServiceEvents to resolve dependencies, and
>> its Service properties to cut down on global datastore space.
>>
>> - Building protocol handlers around a more general-purpose interface,
>> so that they can be used elseproject, then wrapping bundles around
>> them to make them standard services in a Felix environment for
>> Chatterbot.
>>
>> I think Listener and Sender have to remain, rebuilt as services.
>> Changes to make Parser a service should leave the parse() method
>> functionally unchanged.
>>
>> The global datastore (I call it the "namespace" in the proposal; I see
>> now that that conflicts with a term of art) would work best as a
>> service.  I'd like to discuss the Chatterbot Listable class vs. the
>> standard Dictionary or HashTable classes; Listable allows access to
>> subsets of the datastore by using a partial key.
>>
>> So where do I go from here?  A new proposal draft?
>>
>> Don
>>
>> On Fri, Jan 29, 2010 at 11:05 AM, Richard S. Hall <heavy@ungoverned.org>
>> wrote:
>> > On 1/29/10 10:38, Donald Whytock wrote:
>> >>
>> >> I have an overview of the current Chatterbot architecture at
>> >>
>> >> http://www.imtower.net/chatterbot/doku.php?id=overview
>> >>
>> >> Chatterbot is different from JMS inasmuch as it's currently built to
>> >> receive messages from chat IDs and turn them into messages from
>> >> Chatterbot-internal IDs, and vice versa.  My intent was to allow
>> >> multiple chat IDs (same protocol or different protocols) to translate
>> >> into a single Chatterbot ID, so that a user could choose how he
>> >> accessed the bot.  Which protocol a message comes in over should be
>> >> totally transparent to the Parsers, and the Parsers should be able to
>> >> send messages out using Chatterbot IDs and not worry what protocol is
>> >> used to deliver them.
>> >>
>> >> Looking briefly over Felix (http://felix.apache.org), I'd say the
>> >> Chatterbot Listener and Parsers would be equivalent to the Felix Shell
>> >> and Commands, if the Shell was fed a JMS stream consolidated from
>> >> multiple message streams, and if its output was then dispersed over
>> >> multple message streams.  Though there would also need to be a way to
>> >> set up a Command to respond to any input string, rather than one
>> >> starting with a particular word.
>> >>
>> >
>> > Just to be clear, there are two shells at Felix:
>> >
>> >    http://felix.apache.org/site/apache-felix-shell.html
>> >
>> > And
>> >
>> >    http://felix.apache.org/site/apache-felix-gogo.html
>> >
>> > Although they basically do the same thing, I think Christopher was
>> referring
>> > to the latter shell, which is more sophisticated than the former and may
>> > eventually become and OSGi standard.
>> >
>> > -> richard
>> >
>> >> Chatterbot Parsers also have the capacity to originate messages to
>> >> users other than the one whose message the Parsers are responding to,
>> >> so that they can serve as chatrooms; this would be the equivalent of
>> >> Felix sending out notifications to other users when a given user
>> >> performed a command.  Would this compare to Felix Event Admin?
>> >>
>> >> That pretty much just leaves the global namespace, which is
>> >> essentially volatile JDO.  This is where the Chatterbot IDs are stored
>> >> and the Modules are defined; it gets updated by Channels, and can be
>> >> referenced and updated by Parsers.  I've implemented that as a TreeSet
>> >> of TreeSets, due to the key structure, but of course the internal
>> >> structure of the namespace is largely transparent to the modules.
>> >>
>> >> So all in all I'd say there's no inherent barrier to building
>> >> Chatterbot with Felix.  Especially if it'll still run off my USB
>> >> drive.
>> >>
>> >> Don
>> >>
>> >> On Fri, Jan 29, 2010 at 3:44 AM, Christopher Brind<brindy@brindy.org.uk
>> >
>> >>  wrote:
>> >>
>> >>>
>> >>> Hi,
>> >>>
>> >>> I have read through the proposal and I like the idea of it.
>> >>>
>> >>> The only issues I have are around modularity and shell/console.  Apache
>> >>> already has a modularity solution (Felix) based on an open standard
>> >>> (OSGi) I
>> >>> don't think the Java community as a whole needs yet another modularity
>> >>> solution. =)   Felix also provides a shell which allows you to manage
>> >>> module
>> >>> (bundle) lifecycle (install, start, stop, update, uninstall) and while
>> I
>> >>> don't know what the status is regarding the 'Standard Shell' (OSGi RFC
>> >>> 132)
>> >>> it is quite easy to add new commands to the Felix shell.   Felix is
>> also
>> >>> very lightweight, so it wouldn't add much to your footprint, but would
>> >>> give
>> >>> you a sophisticated dynamic module contain in which to work as well
as
>> >>> making it compatible with various environments already using OSGi now
>> >>> (e.g.
>> >>> Application Servers, etc).
>> >>>
>> >>> I could see potential uses for this project in my own work, but as I've
>> >>> implied, it would have to be compatible with OSGi which is where I
>> spend
>> >>> most of my time.  I'd even offer to assist that effort on this project.
>> >>>
>> >>> This is more of a question, is there any Java API standard abstraction
>> >>> for
>> >>> chat protocols?  e.g. javax.chat?  I don't think there is but there
is
>> of
>> >>> course JMS, is ChatterBot significantly different from JMS?  If so,
>> >>> perhaps
>> >>> a low priority side goal of the project should be to develop a standard
>> >>> Java
>> >>> API standardisation for chat?
>> >>>
>> >>> Cheers,
>> >>> Chris
>> >>>
>> >>>
>> >>> On 29 January 2010 03:32, Donald Whytock<dwhytock@gmail.com>  wrote:
>> >>>
>> >>>
>> >>>>
>> >>>> Hello all...
>> >>>>
>> >>>> As discussed before, here is the current wiki text of the proposal
for
>> >>>> Chatterbot, a lightweight framework for chat responders.  The proposal
>> >>>> is at
>> >>>>
>> >>>> http://wiki.apache.org/incubator/ChatterbotProposal
>> >>>>
>> >>>> Interested in comments, feedback and participation.
>> >>>>
>> >>>> Thanks...
>> >>>>
>> >>>> Don
>> >>>>
>> >>>> - wiki text -
>> >>>>
>> >>>> Abstract
>> >>>>
>> >>>> ChatterBot is a lightweight, multiprotocol framework and container
for
>> >>>> chat responders.
>> >>>>
>> >>>> Proposal
>> >>>>
>> >>>> ChatterBot is a framework for developing chat responders (applications
>> >>>> that respond to messages received via online chat) and a container
for
>> >>>> deploying them. It is written in Java SE, and runs as a Java
>> >>>> application. Chat responders are built by extending a single class
and
>> >>>> modifying a configuration file to reference the new class.
>> >>>> ChatterBot's focus is on the following characteristics:
>> >>>>
>> >>>> - Small: The current framework consists of eight core classes.
>> >>>>
>> >>>> - Standalone: ChatterBot does not require external servers to operate.
>> >>>>
>> >>>> - Portable: ChatterBot should work as run from any Java-capable
>> >>>> machine. For full functionality that machine should have internet
>> >>>> access, but localhost and console connectivity are possible. It
should
>> >>>> be possible to run multiple instances of ChatterBot on the same
>> >>>> machine or on separate machines with no loss of functionality.
>> >>>>
>> >>>> - Extensible: An instance of ChatterBot can support multiple message
>> >>>> parsers and protocols. Adding more is done by editing a configuration
>> >>>> file.
>> >>>>
>> >>>> - Dynamic: Activating and de-activating modules should be possible
>> >>>> during runtime.
>> >>>> Multi-user access: Multiple users, over multiple protocols, should
be
>> >>>> able to access deployed applications.
>> >>>>
>> >>>> Rationale
>> >>>>
>> >>>> A chat responder can serve as a user interface to applications,
either
>> >>>> those built into the responder or external applications with which
the
>> >>>> responder communicates. Such an interface is more secure than
>> >>>> interfaces such as Telnet or web services since it does not require
>> >>>> open ports in the firewall; the container connects out through the
>> >>>> firewall to the chat server, rather than allowing users to connect
in.
>> >>>>
>> >>>> A lightweight chat responder can be installed on any system to allow
>> >>>> command-line access to users over whatever protocol a user may have
>> >>>> access to. Thus applications can be accessed from web interfaces,
>> >>>> instant-message systems, text messages, email, etc. A scalable
>> >>>> container can allow as many or as few access protocols as are desired.
>> >>>>
>> >>>> ChatterBot, therefore, has value for those circumstances where a
user
>> >>>> interface is needed but a server-based or enterprise solution is
>> >>>> either not possible or not desired. It also can serve as a bridge
>> >>>> between applications, where one or more uses a chat protocol such
as
>> >>>> XMPP to communicate.
>> >>>>
>> >>>> Background
>> >>>>
>> >>>> ChatterBot began in 2005 as a thin-server approach to online
>> >>>> multi-user board games, implemented as applets sending gamestate
>> >>>> changes to one another via message relaying. The idea was to make
as
>> >>>> general-purpose a server as possible, so that multiple games could
be
>> >>>> built that employed the same message-relaying system.
>> >>>>
>> >>>> Version 0.2 of the server was then refined in 2008 to allow for
more
>> >>>> varied and functional message-handlers, and was used to implement
a
>> >>>> room system that allowed for room-specific applications -- parsers
>> >>>> that checked the user's room before handling a command and sent
>> >>>> responses to other room occupants. This version was structured
>> >>>> entirely around XMPP. The global namespace was introduced to allow
>> >>>> modules to communicate with relatively limited coupling.
>> >>>>
>> >>>> The current version, 0.3, as of late 2009, functions with XMPP and
has
>> >>>> the capacity to function with whatever other protocols channels
are
>> >>>> coded for. Applications built using 0.2 are being ported to 0.3.
At
>> >>>> this point the original thin-server backend intended in 0.1 would
be
>> >>>> built as an application using 0.3.
>> >>>>
>> >>>> Current Status
>> >>>>
>> >>>> Meritocracy
>> >>>>
>> >>>> Peer review and alternate ideas are welcomed in this project with
open
>> >>>> arms. This project was intended specifically as an alternative to
>> >>>> traditional server-based or enterprise architecture; however, it
is
>> >>>> recognized that tried-and-tested principles established in enterprise
>> >>>> architecture may be applicable here.
>> >>>>
>> >>>> Core Developers
>> >>>>
>> >>>> As of late 2009, there is one developer, Donald Whytock (dwhytock
at
>> >>>> gmail dot com). Donald Whytock has several years of experience as
a
>> >>>> software developer, working in a variety of languages, including
C,
>> >>>> Java, Perl, PHP, JavaScript and SQL. He develops both professionally
>> >>>> and casually; ChatterBot has been an independent, voluntary effort.
>> >>>>
>> >>>> Alignment
>> >>>>
>> >>>> ChatterBot's primary potential alignment with ASF is that of a
>> >>>> framework for internet-accessible applications. At its core, it
is
>> >>>> largely free of outside dependencies, though modules can be built
to
>> >>>> utilize other technologies. Embedded Derby is used in one application;
>> >>>> use of Derby and/or ORM should be explored as a base capability.
>> >>>>
>> >>>> Initial Goals
>> >>>>
>> >>>> ChatterBot currently exists as a functioning prototype; protocol
>> >>>> modules built for it provide access to chat responders via
>> >>>> XMPP/Jabber, localhost connections and a chat-simulating console.
>> >>>> Further development is to consist of refinement of the core classes
>> >>>> and expansion of the secondary modules.
>> >>>>
>> >>>> Core Classes
>> >>>>
>> >>>> Shell: The main-method class, used to launch the application.
>> >>>> Potential refinements: re-entrance, clean shutdown, restart
>> >>>>
>> >>>> Listable: The foundation class for the global namespace.
>> >>>> Potential refinements: configuration file format, persistence
>> >>>>
>> >>>> Module: The interface for all modules loaded by Shell.
>> >>>> Potential refinements: restart, shutdown
>> >>>>
>> >>>> Channel: The interface for protocol handlers that accept incoming
and
>> >>>> outgoing messages.
>> >>>> Potential refinements: an interface for relaying XML/HTML data within
>> >>>> messages
>> >>>>
>> >>>> Listener: The driving module that routes messages to Parsers.
>> >>>> Maintains a list of Parsers, submitting an incoming message to each
>> >>>> Parser in turn until a Parser indicates successful handling.
>> >>>>
>> >>>> Parser: The abstract class for message-parsing modules.
>> >>>> Potential refinements: built-in parsing/tokenization, persistence
>> >>>>
>> >>>> Sender: The module that routes outbound messages from Parsers to
>> >>>> Channels.
>> >>>> Potential refinements: time-delayed messages, in-system messages
>> >>>>
>> >>>> Secondary Modules
>> >>>>
>> >>>> XMPPChannel: Extends Channel; protocol handler for XMPP.
>> >>>>
>> >>>> LocalhostChannel: Extends Channel; handler for localhost connections
>> >>>> with other processes.
>> >>>>
>> >>>> ConsoleChannel: Extends Channel; supplies a simple Swing console
for
>> >>>> entering messages and receiving responses.
>> >>>>
>> >>>> INIParser: Extends Parser, allows examination and manipulation of
the
>> >>>> global namespace.
>> >>>>
>> >>>> New modules should be developed to add optional functionality. In
>> >>>> particular, new Channels should be developed for AIM, YM, MSN, etc.
>> >>>> Other potential modules include:
>> >>>>
>> >>>> SystemParser: Extends Parser, allows dynamic activation and
>> >>>> de-activation of modules.
>> >>>>
>> >>>> FileXferParser: Extends Parser; implements file transfer between
>> >>>> client and ChatterBot's host. Will require refinement of Channel
and
>> >>>> protocol-specific extensions of Channel.
>> >>>>
>> >>>> DB: A database interface. One application built using ChatterBot
>> >>>> currently uses embedded Derby as its interface, preserving server
>> >>>> non-dependence.
>> >>>>
>> >>>> RoomParser: Extends Parser; implements chatrooms, relaying messages
>> >>>> among users in a room and allowing room-specific applications.
>> >>>>
>> >>>> Known Risks
>> >>>>
>> >>>> Orphaned Products
>> >>>>
>> >>>> Currently the project has only one committer, though Donald Whytock
>> >>>> has been working on the code for a few years and is committed to
>> >>>> seeing a functional product available.
>> >>>>
>> >>>> Inexperience with Open Source
>> >>>>
>> >>>> While the developer has experience working with open-source products,
>> >>>> this is the first time opening up a project for open-source
>> >>>> collaboration. As modular as the project is, however, open-source
>> >>>> collaboration should not be a problem. It is greatly desired that
this
>> >>>> project not be developed in a vacuum.
>> >>>>
>> >>>> Fascination with the Apache Brand
>> >>>>
>> >>>> Association with the Apache brand is not sought for personal
>> >>>> publicity; rather, it is sought for the associated community and
>> >>>> access to collaboration and peer review. This project will see its
>> >>>> full potential through public use and refinement, and a product
more
>> >>>> refined for everyone's use is a more refined product for the
>> >>>> developer's use as well.
>> >>>>
>> >>>> Initial Source
>> >>>>
>> >>>> Original code developed by Donald Whytock.
>> >>>>
>> >>>> Required Resources
>> >>>>
>> >>>> Mailing Lists
>> >>>>
>> >>>> chatterbot-private
>> >>>> chatterbot-dev
>> >>>>
>> >>>> Subversion Directory
>> >>>>
>> >>>> https://svn.apache.org/repos/asf/incubator/chatterbot
>> >>>>
>> >>>> Issue Tracking
>> >>>>
>> >>>> JIRA ChatterBot
>> >>>>
>> >>>> Initial Committers
>> >>>>
>> >>>> Donald Whytock (dwhytock at gmail dot com)
>> >>>>
>> >>>> ---------------------------------------------------------------------
>> >>>> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>> >>>> For additional commands, e-mail: general-help@incubator.apache.org
>> >>>>
>> >>>>
>> >>>>
>> >>>
>> >>>
>> >>
>> >> ---------------------------------------------------------------------
>> >> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>> >> For additional commands, e-mail: general-help@incubator.apache.org
>> >>
>> >>
>> >
>> > ---------------------------------------------------------------------
>> > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>> > For additional commands, e-mail: general-help@incubator.apache.org
>> >
>> >
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>> For additional commands, e-mail: general-help@incubator.apache.org
>>
>>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Mime
View raw message