Tuesday, September 27, 2005
yahoo widget presentation
Talk by Dan Theurer, Technical Evangelist at yahoo
Started with power of yahoo APIs of:
1. MAP : how one can overlay his/her application data on top of a yahoo map, embed those applications etc.
Also digressed to advantages of yahoo over google:
a) Yahoo does not limit the number of hits to a low level; instead 5000 queries a day or so
2. Yahoo music engine: access to music engine object model using javasrcript
3. Search API: advanced we interface
4. "Context analyisis" for yahoo searches
----
Konfabulator talk:
Control statements, UI controls, timers triggers framework, "widget convertor" to package zip file.
Way of arranging data under widgets: widget folder \WidgetName at the top level
"On load" trigger example was shown.
Also supports debug mode with tag; windows cosnole
Check http://www.widgetgallery.com
and
http://developer.yahoo.net/community
Started with power of yahoo APIs of:
1. MAP : how one can overlay his/her application data on top of a yahoo map, embed those applications etc.
Also digressed to advantages of yahoo over google:
a) Yahoo does not limit the number of hits to a low level; instead 5000 queries a day or so
2. Yahoo music engine: access to music engine object model using javasrcript
3. Search API: advanced we interface
4. "Context analyisis" for yahoo searches
----
Konfabulator talk:
Control statements, UI controls, timers triggers framework, "widget convertor" to package zip file.
Way of arranging data under widgets: widget folder \WidgetName at the top level
"On load" trigger example was shown.
Also supports debug mode with
Check http://www.widgetgallery.com
and
http://developer.yahoo.net/community
Tuesday, September 20, 2005
Event Server presentation
Presentation from ispheres: http://www.ispheres.com; the VP of engineering Kal Krishnan gave a nice presentation, focussed on the event server piece. The core strength of the event server are:
1. EPL or event processing language with (IF-WHEN-THEN) syntax. This language is 'SQL' like but also maps very closely to java and compiles into java byte code.
2. Support both pull and push model for obtaining data from different disparate sources
3. Support of TIME_WINDOW(), LENGTH_WINDOW() constructs to detect events. In effect, this allows for filters to be built on moving averages, time constraints, even regression analysis results
Asked for reasons of inventing EPL: expressivity of above constructs not possible with straightforward SQL. There does exist support of persistence in the framework as well. The concept talked about three different streams:
Raw input ---(Stream A filter)-----> Analyze data---(Stream B filter)---->Conclude---(Stream C filter)---->Output
The company claims high performance, scalability and extensibility
Perf: cannot be verified since no standard benchmark program.
Scalability: Can be put on more powerful CPU or distribued across machines
Extensibility: different data sources, including even fields of a java bean object.
There are competing products in the market, including 'event SQL' framework from stonebreaker, another IBM effort as well (not known?)
There is effort on standardization: CEP or Complex events processing group; lot of potential for technology now and future.
My two ideas:
1. RFID space
2. Telecom perf and alarm mgmt
1. EPL or event processing language with (IF-WHEN-THEN) syntax. This language is 'SQL' like but also maps very closely to java and compiles into java byte code.
2. Support both pull and push model for obtaining data from different disparate sources
3. Support of TIME_WINDOW(), LENGTH_WINDOW() constructs to detect events. In effect, this allows for filters to be built on moving averages, time constraints, even regression analysis results
Asked for reasons of inventing EPL: expressivity of above constructs not possible with straightforward SQL. There does exist support of persistence in the framework as well. The concept talked about three different streams:
Raw input ---(Stream A filter)-----> Analyze data---(Stream B filter)---->Conclude---(Stream C filter)---->Output
The company claims high performance, scalability and extensibility
Perf: cannot be verified since no standard benchmark program.
Scalability: Can be put on more powerful CPU or distribued across machines
Extensibility: different data sources, including even fields of a java bean object.
There are competing products in the market, including 'event SQL' framework from stonebreaker, another IBM effort as well (not known?)
There is effort on standardization: CEP or Complex events processing group; lot of potential for technology now and future.
My two ideas:
1. RFID space
2. Telecom perf and alarm mgmt
