Ted Patrick > { Events & Community } > Adobe Systems

Don't let your end users get the better of you...\0



User Obliteration
Distributing Clue to Users
Conducting Black Operations in the IT Theater

And much more...

Happy Halloween!

Cheers,

ted ;)

Yes, I realize that MX2004 just shipped (and soon again) but I would argue that now is the perfect time to suggest new features.\0

I try to think in terms of Macromedia's product development cycle. You see when we are learning the new beast, they have moved onto development of the next one. Like all software projects they start with high level planning and a broad requirements discussion. What is truly amazing about Macromedia is that our opinion really does shape product development directly. I never realized this until I experienced it first hand.

Just after the release of Flash MX, it was January 2002. I was complaining on Flashcoders, to "Fork the IDE" in that the Flash product was serving too many masters (designer, animator, developer) yet didn't fully satisfy any one market completely. I then received an email from Peter Goldie (Macromedia) and was asked to have a conversation about Flash and PowerSDK. Peter wanted to know exactly what I wanted in the product and how it would better suit my needs and the needs of my clients. At the time, I wanted a specialized IDE for component development (truth be told, I still do). I complained that the development paradigm was not ideal for developing applications with components and that for some the timeline paradigm wasn't a good fit. The discussion also leaned toward accommodating Delphi and VB users with a consistent event/component model. The discussion ended and lots of time passed.....

18 months later.....MX2004 & MX2004 Professional Released

The first thing that caught my eye was that Flash was now 2 distinct products. One was specialized for the development of applications and that version supported a paradigm that didn't have a timeline, Form based development. The component model had been overhauled and the event model used single words lacking the preceding "on". I then recalled my conversation with Peter, wow they really listened!

Now please don't hate me all at once, I was not directly involved in the addition of these features, the heavy lifting was done by much more talented individuals. I cannot imagine delivering 2 development paradigms in a single product. Actually I would wager that Royale adds yet another paradigm to the Flash Professional IDE. ;)

What is very important here is that:

1. Macromedia listens intently to their customers.
2. Macromedia acts based on our opinion of what a product should do.
3. Now is the perfect time to affect change within the Flash product line.

If you have a feature itch, send it to Macromedia, post it to Flashcoders (OT) or Blog about it. You might be pleasantly surprised when you open MX 2006.

Here is my request.

Cheers,

Ted ;)

Continuations are a simple threading model that is supported in several stack based systems. So what exactly are continuations? And how should Flash support them?\0

Continuations are special functions that never return to the stack and allow the stack to continue processing. The execution of a continuation is thus parallel to the operation of the stack. Continuations run until they complete. Most often they execute another continuation or call themselves recursively. When execution reaches the end of a continuation, the execution ends and all resources are returned / garbage collected.

In most stack based systems (Flash, Python, Java), when you call a function, stack processing moves to processing the function and returns to the next instruction after the function is executed. This forces the stack to always work under a single thread and never allows the system to process 'N' items in parallel.

Example of continuation support for ActionScript:

myContinuation = continuation(){

trace('myContinuation Executing')

//if data is true call the continuation again
if (mydata ==1){ arguments.callee() }

//if this point is reached kill the continuation
}

mydata = 1
myContinuation()
trace('hello')
mydata = 0
trace('end')

Output:
myContinuation Executing
hello
myContinuation Executing
end

Output without continuation support:
myContinuation Executing
256 levels of recursion were exceeded in one action list.
This is probably an infinite loop.
Further execution of actions has been disabled in this movie.

This would add a simple light parallel execution model to the Flash Player providing a solution to a entire host of problems:

1. Parallel instantiation - Shorten the delay in instantiating components, frameworks, etc.
2. Parallel execution of event listeners - All events would be processed at close to the same time frame, vs. sequentially.
3. Parallel events execution - Allowing player events to be processed at once independently.
4. Process XML in parallel
5. {Your reason here}

A system independent threading model is just what the Flash Player needs to escape the performance related problems of the current architecture. Even with the enhancements in v7 of the player, the V2 component model takes the player to the performance limit. This is a direct result of Flash being a stack based system without support for parallel processing.

To support this initiative please do the following:

1. Sign the comments showing your approval for Macromedia to develop this feature
2. Send this request in parallel to Macromedia Wish Form

Comments and suggestions are always welcome.

Thanks,

Ted Patrick
PowerSDK Software Corp.

I need your help. We are building a specialized chat server for Flash and wanted to release the Alpha API for public comment. An alpha version of the server will be released next week allowing you to run the server locally and build applications. Any feedback you can provide would be most helpful to me. Please forward comments to ted@powersdk.com.

Thanks,

Ted :)

PSDK:Chat Client API - Alpha Review
-------------------------------------------------------------

// make a server object
myChat = new com.psdk.server.chat()

// connect to a server:port
myChat.connect ('server', 7000)

//login to the server with a username
myChat.login('ted')

There are 2 primary objects, users and rooms. Once you login, these objects appear, populate automatically, and allow you to send and receive messages. Here are a few ways to use the users and rooms objects:

// send a message to 'Fred'
myChat.users.fred.send('Yo what is up!')

// execute a method on freds client
myChat.users.fred.explode('abc',123)

// get info on fred
myChat.users.fred.info()

//Loop through users and send a message to all
for (var a in myChat.users){
myChat.users[a].send('Weeeeeee...')
}

// enter a room - if the room doesn't exist, it will be created automagically
myChat.rooms.hotel.enter()

// send a message to the room
myChat.rooms.hotel.send('This room is great!')

// execute a room method - This goes to all users in the room and executes on the room object
myChat.rooms.hotel.explode('KaBoom!')

// get info on hotel
myChat.rooms.hotel.info()

// exit a room - if you are the last one in the room, it is destroyed automagically
myChat.rooms.hotel.enter()

//Loop through rooms
for (var a in myChat.rooms){
trace(a)
}

When you receive a message, events are fired in several locations depending on the object in question.

1. Add an event to a class

//class for user objects
com.psdk.server.chat.user.prototype.onlogoff = function(){}

//class for room objects
com.psdk.server.chat.room.prototype.onrecieve = function(message){}

2. Add an event to the Chat instance - All events are recieved here
myChat.onlogin = function(){}

3. Add an event to a User or a Room instance

myChat.users.ted.onrecieve = function(message){}
myChat.users.ted.onrecieve = function(message){}

4. Subscribe to events via a listener
myChat.addListener('onclose', this)

Chat instance - onconnect, onclose, onlogin, onlogoff, onenter, onexit, oninfo, onexecute, onrecieve, onerror
Room instance - onenter, onexit, oninfo, onexecute, onrecieve, onerror
User instance - onlogin, onlogoff, onenter, onexit, oninfo, onexecute, onrecieve, onerror
Listeners - onconnect, onclose, onlogin, onenter, onexit, oninfo, onexecute, onrecieve, onerror

The current build of the library adds 2.4Kb to your swf for full functionality. The library will also ship in .as and swf formats to allow runtime importing and use of the server. The client will be open source using a BSD license

About the server:

The server uses a special optimized core that allows us to support many types of socket based protocols including, HTTP, HTTPS, Remoting, RTMP, XMLSockets in a scalable cross platform manner. The core was built independent of the individual servers and allows us to create servers on demand. The individual servers are simply events that render a protocol back to a client via a socket. It is very simple design that allows us to build sophisticated servers in a fraction of the time.

The server will be supported on the following platforms: OSX, Win32, Linux, FreeBSD, BSD. In intial testing, the core was able to support an abusive quantity of connections as follows: Windows:2200 connections, FreeBSD:8000 connections, OSX:6000 connections, Linux:untested. An advanced model of the server will support clustering across many servers.

On the chat server, the protocol is plain text (not xml) thus the messages are very small as methods are single characters space delimited with null byte termination. The chat server does not store any data on the local file system and all processing is in memory making user and room processing very fast.

The price point for the chat server is very low compared to the current market as we will be providing customization services allowing us to tailor the server protocol and client library to a specific application. The default chat server will sell for below $150 and there are no connection / user restrictions. Clustering support will be additional in a later version. Development servers will be free and are restricted to running on the localhost.

Again, any thoughts on the Client API would be most helpful and depending on the feedback might be worth of a free license.

Thanks,

Theodore Patrick - Founder
PowerSDK Software Corp.

The debate over AS1 or AS2 is interesting but in the end its all bytecode. Regardless of what toolset made the swf file, the player will interpret the bytecode instructions(compatible or not). What is important is that the swf meets your requirements for functionality and usability.

Different development approaches have merit but all approaches are not created equally for the problem at hand. Some problems are better solved procedurally, some via objects, some via components, and some via a framework or a pattern. In the end you are trying to solve a problem and the toolset simply provides options towards a particular solution. The argument that there is only one way to solve a problem is ego-driven crap.

Personally, I am very comfortable with AS1 and can solve problems with it easily. Having worked with it for years, I know the ins and outs of shaping solutions with it. When I write a line of code with AS1, I know what will happen at runtime. With AS2, that isn't the case, yet.

AS2 provides compile time tools to allow you to detect subtle errors in development. The additions of Classes, Interfaces, and Type Safety are very powerful if used properly and very destructive if used improperly. Developing with them effectively requires understanding how they impact software at runtime. Unfortunately there are only a handful of developers that understand the runtime perspective of AS2 (most work for MM), although this will change over time.

If you understand what a language does at runtime, you can use it in many unexpected ways to solve problems. Good developers understand the runtime perspective of a language, great developers shape solutions from this understanding.

"I should think you Jedi have more respect for the difference between knowledge and wisdom." -Dex

cheers,

ted ;)

I find it interesting that Apple continues to update the X86 version of OSX. One could say that they are keeping their options open in regard to future chipsets, but selling a first class hardware compatible alternative to Windows is more like it. Lets face it they are in the business of making money and the x86 market is too large to ignore.

I have been a FreeBSD user for over 6 years and love it as a server operating system, but I just refuse to use it on the desktop because I lack the apps. Instead I continue to use windows simply due to my legacy investment in software. Although with MACR DEVNET, I may switch to OSX when my I renew my license.

I recently deployed a project using FreeBSD. We clustered 3 servers running Jabber, 2 servers handle socket connections and the other server handles Jabber protocol and the user database. Currently the system supports 16,000 concurrent users as the socket servers can maintain 8,000 connections apiece using kqueue events on FreeBSD. The system is easy to scale as we simply add more socket servers to the cluster and add them into the load balancer. The 3 servers have been running for 10 months straight without a reboot, you sure can't do that with windows!

Personally I think apple will release to x86 as soon as OSX achieves critical mass for applications. They cannot remain in a small niche OS market forever and the x86 architecture is in desperate need of a solid visual operating system.

cheers,

ted ;)

It seems that you can seamlessly integrate stored procedure calls with ActionScript Remoting. At first I was a bit skeptical that this would work, but some engineer at MM knew what they were doing. Here is a simple ASR remoting function using a stored procedure:

function getsystem(o){

// create a stored procedure call as a string
var sql = 'sp_GetVersionInfo @id=' + o.get('id')

//execute the data via CF.query()
var ret = CF.query("data001",sql)

//return the sql result set to remoting
return ret

}

We are using SQL Server 2000 and ColdFusion MX. The database 'data001' is defined in the CF admin. Passing variables by name is a snap. Simply use the @{name} and use single quotes with stings (same as sql) and your all set. This makes using stored procedures easy.

Special thanks to Dave Gallerizzo at Figleaf for working through fine details to get this working.

Cheers,

ted ;)

I installed Google Toolbar in my browser and suddenly a majority of Flash sites don't work! \0

When you designed your site, you put the core of your content into a pop-up and now I can't see it without pressing CTRL usually on the 3rd or 4th time! Actually I don't even realize that you have content on your site unless I catch the cursor change denoting a blocked site.

Google Toolbar - Pop-Up Blocker

Please switch your site from JavaScript POP-UP to Inline Flash.

Cheers,

Ted ;)

Although XMLSocket implies the use of XML, it is not required. The key to XMLSocket messaging is delimiting using Null bytes or '\0'. Using this technique you can avoid the xml parsing delay and use a string or binary encoding over an XMLSocket. The results are simply fantastic and speeds up system performance by a factor of 5 at a minimum for both client and server. For game use this is essential.

This requires a server that understands the string / binary encoding, here is a small socket server implementation in python:

import socket
HOST = '' # Symbolic name meaning the local host
PORT = 5222 # Arbitrary non-privileged port
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.bind((HOST, PORT))
s.listen(1)
conn, addr = s.accept()
print 'Connected by', addr
while 1:
data = conn.recv(1024)
data = data[0:len(data)-1]
if not data: break
print "Data received:",data," ",len(data)
if data == 'foo':
print "Data sent:bar"
conn.send('bar\0')
break

On the client side, you use the following to avoid the XML parser:

x = new XMLSocket()
x.onConnect = function(success){
trace('onConnect '+success)
x.send('foo')
}
x.onData = function(x){
trace('onData '+x)
if(x == 'bar'){ trace('bar message!')
}
x.connect('localhost',5222)

The key is in using the onData event vs the onXML event. onData receives a string and onXML receives an XML object. Using onData avoids the XML parser on the client allowing you to use simple string parsing to execute a named function with arguments.

On a current project we are using this technique to render information to screen. The encoding pattern is very simple and supports 3 datatypes(number, string, hex) using space as a delimiter. These values are passed as follows:

x.send('onglyph 23 34 x394393 donut')

The server receives:
'onglyph 23 34 x394393 donut\0'

This is translated to a function call:

onglyph(23,34, 'x394393' , 'donut' )

The results of this function are then returned to the client in the same encoded format.

Although we don't use XML as a whole, that doesn't exclude it from being used as an argument. As XML is just a string, any argument in this encoding can easily be replaced by a XML document. Win, Win for structured data! ;)

Especially in regard to game use, this format is very small, very fast, and very efficient on both the client and server. We have seen great improvements in the performance by using this format. It is also very bandwidth friendly as the message size shrinks dramatically without XML encoding.

Enjoy,

Ted ;)

Special thanks to Moises for pointing out SnorComments that work well with Blogger. You sure can't beat free.

Cheers,

ted ;)

I have refined my Eolas solution for Flash a bit. Specifically it has now "PASSED" testing with the test version of Internet Exploder.

I would like to stress that this solution might be non-compliant as the changes to IE are not firm and currently represent "beta / alpha" code. The solution was tailored to the language of the proposed changes and not to a hole in the implementation. Here is a snip from Microsoft's site on the subject:

http://msdn.microsoft.com/ieupdate/activexchanges.asp

Affected Web Pages
All Web pages with ActiveX controls are affected by this change unless they meet one of the following conditions.

- The controls are created dynamically from script loaded from another location.
- The controls do not reference remote data.

Avoiding the first case, you can instantiate an OBJECT with minimal information other than the OBJECT tags. Since this doesn't reference remote data initially, the OBJECT doesn't warn the user. Next we modify the OBJECT with JavaScript using the ID that the OBJECT was instantiated with and VIOLA the object is working properly.

The key is determining which items in an OBJECT refer to remote data. Personally I feel several portions of the standard OBJECT tag set violate this rule:

1. CODEBASE="http://download.macromedia.com/pub/shockwave/cabs/flash/swflash.cab#version=6,0,0,0"
2. PARAM NAME=movie VALUE='mymovie.swf'

Outside of those two URLS, the OBJECT tag is clean. These items are added into the OBJECT with the JavaScript as follows:

window.{ID OF OBJECT}.codebase="http://download.macromedia.com/pub/shockwave/cabs/flash/swflash.cab#version=6,0,0,0"
window.{ID OF OBJECT}.LoadMovie(0,"sample.swf")

The second tag uses a Flash Player method that has been available since F3 (Version 3 Flash) of the ActiveX control.

Here is a good example:

eolas.zip

Cheers,

ted ;)

The Eolas issue has 2 distinct criteria: 1. The controls are created dynamically from script loaded from another location. 2.The controls do not reference remote data.

Thus far I have only seen 1 recommended class of solution. Another option is potentially to modify object properties after object instantiation, thus working around the MS rules so to speak.

Initially object tags must not reference remote data in the declaration and thus do not warn the user about their use. Next a small bit of JavaScript modifies the object tag as the page loads. We then use a player external method to load data into the player via LoadMovie via JavaScript.

Here is a sample:

http://www.powersdk.com/eolas/index.html

I am unsure of a few specifics:

1. Can you edit object properties after the page loads and still not violate the rules
2. Can you modify object properties to point to external data

In the second item, the Macromedia .CAB URL and the violate this rule, but can be left out if there is a player 4+ supporting the JavaScript loadMovie. In this case the external data is being requested within the player, not from within the Object tag. This workaround also has some advantages as you can pass data to the url.

The JavaScript looks like this:


window.sample.codebase="http://download.macromedia.com/pub/shockwave/cabs/flash/swflash.cab#version=6,0,0,0"
window.sample.width=800
window.sample.height=600
window.sample.LoadMovie(0,"sample.swf")
window.sample.classid="clsid:D27CDB6E-AE6D-11cf-96B8-444553540000"
Not sure if I broke the EOLAS rules, but it works with all current content. In the sample, I removed Netscape embed to isolate the IE only issue. I am not sure if the nested Embed will trigger a warning as is standard practice for Flash authors.

Personally, I wish they would just pay the bill and be done with it. Can you imagine Redmond actually licensing something legally?

Cheers,

ted ;)

I am working on my presentation for max and found something interesting. In V2 component, each component gets an queue for the various events supported. These queues are created when addEventListener is called.

Basically __q_click is the queue for the 'click' event. The naming convention goes as follows:

"__q_" +Event.type >> See mx.core.events.EventDispatcher & UIEventDispatcher

Although what is strange is that __q_click is a function!!! Yes, instead of putting listeners in an array, MM hid then inside a function and __q_click is also protected with ASSetPropFlags so you will not find it looping accross a component.

Here is a sample:

1. ADD a FUI v2 "Button to the stage
2. Name the instance "myComponent" >>> Case Sensitive now!!!
3. Add this code to the main timeline

//create a simple function
myClick = function(){trace('myClick')}
dog = {click:function(){trace('dog.click')}}

//add the function directly as a listeners
myComponent.addEventListener('click', myClick)
myComponent.addEventListener('click', dog)

//trace the queue
trace(myComponent.__q_click)

//loop through the queue and trace the listeners
for (var x in myComponent.__q_click){
trace(x +':'+myComponent.__q_click[x])
}

//execute the listener directly
myComponent.__q_click[0]()

What is weird to notice is that execution order is the reverse of the order the items were added. Basically this is symbolic of using a for loop in an object/function. When you use a for loop, you are looping through the order items are stored in raw memory and thus items added later go first.

This also presents a fairly simple hack to clear all listeners of a certain type:

myComponent.__q_click = function(){}

Although this would be better implemented in as a prototype method. Bad Ted, Bad!

See you at MAX in Novmeber.

ted ;)

There is another serious hole in MSIE 5.01, 5.5, and 6.0. IE allows a Trojan installation of an ActiveX control. The recommended solution has users switching ActiveX to "prompt before running ActiveX controls", and that means Flash.

http://www.technewsworld.com/perl/story/31756.html

Google

I modified my system and it is really annoying, almost every page has Flash on it! Prompt, Prompt, Prompt, every web page.

Here is the quote from MS:

"You can help protect against this vulnerability by changing your settings for the Internet security zone to prompt before running ActiveX components," Microsoft said.

Terrible news for the business of Flash, the deployment of Flash, and recommending Flash to consulting clients.

Not good. Not good.

Ted ;(

I just finished a 10 month project writing a component based Jabber client in Flash complete with talking characters and all that jazz. So I am taking the day off and going diving in Port Antonio, Jamaica. Ok, it sounds exotic but I live in Kingston, Jamaica and Port Antonio or Porti, as its locally know, is just 2 hours away.




I know everyone has heard of Montego Bay, Ocho Rios, and Negril, but Porti is 10 times better. First it is small and secluded on the north east coast of Jamaica. Second, tourists do not know it exists except for some well informed Europeans that frequent the place. Third, it is a hidden diving gem.

Porti has some of the best reefs in the Caribbean, hard to believe isn't it! The problem is that Jamaica is so large that it rains daily inland and this runoff makes the visibility a touch lower than perfect. Funny thing is in the dead of Hurricane season, it hasn't rained allot so the visibility is now over 100 feet. Now you won't find lots of dive shops, actually there is just 1. On my many visits, we have had frequent encounters with all sorts of sea life, green turtles, sharks, dolphins, rays, actually I saw a spotted eagle ray on my last visit. Hopefully we can find it again, they tend to frequent an identical spot. I guess the main attraction is the soft coral. Due to the heavy mineral content of the water(rainfall runoff), the sponges are really large. Some are easily bigger than me and I am not a small guy.






Well I am off to start my trip and I am taking my dive camera. And yes, I took all the pictures on the page. ;)

It is hard to imagine that after 10 months this project has finally ended and better yet today I will be enjoying diving in Porti!

Cheers,

ted ;)

Remoting is limiting if you don't provide a custom target. In most remoting cases, you call a remoting method 'event.foo()' and data is returned to 'event.foo_result(n)'. Ideally you pass a target with your request and that target is used to fire a specific method. Here is a sample:

ASR:

function sqlTargetExecute(o){
//store the passed sql value
var sql = o.get('sql')

// get the dataset from the database
var rds = CF.query("data001",sql)

//create a hashmap > Equivalent to Flash Object
var ret = java.util.HashMap()

//add the target property passed to the hashmap
ret.put("target", o.get('target'))

//add the resultSet to the hashmap
ret.put("data", rds)

//return it to remoting
return ret
}

Client AS:

#include "netservices.as"
conn = netservices.creategatewayconnection()
event = conn.getservice("event.event2010", this)
sqlTargetExecute = function(sql, target){

// build an object to pass to remoting
// I prefer using a single object to allow for reuse across many functions
// This is especially helpful when you nest functions in SSAS
var o = {sql:sql, target:target}
//event is the remoting object
event.sqlTargetExecute (o)
}
sqlTargetExecute_result = function(n){
// fire the result data at the passed target
[n.target](n.data)
}

sqlTargetExecute('select * from donut','_level0.ondonut')
_level0.ondonut = function(n){
trace('OnDonut has the data!')
}

Actually this makes remoting highly reusable on both the server and client side. Typically using this approach, you only need 1-2 SS or CS methods since any object can call remoting and get data sent to any path in the player. This is also ideal for when you use loadMovie to bring in external reports that need to access remoting. The reports are swf files containing display components and simple functions calling remoting methods.

On my designating SQL on the client side, that is a bad practice for public sites. Use with caution, bad Ted, Bad!!!

Cheers,

Ted ;)