Any meaningful javascript application is going to make use of some kind of application API that lives off on a server somewhere. Generally, some user interface triggers some event which, in turn, will send a request to a remote API. These requests are generally asynchronous. That is, the request happens in the background while the user interface is still responding to events. The number of API connections and requests can add up rather quickly. Even if the data sent over the network for each request and corresponding response are relatively small, there is still the overhead involved with assembling the request.
Using a transactional javascript API access pattern can help cut down on network bandwidth. This is done by minimizing the work done by the API on a given request. Additionally, we also cut down the response size for a given request. The bandwidth consumption will grow to a much larger size if each API response is sent to the javascript application in a single HTTP response. It would be nice if a single HTTP response could contain multiple API responses. One potential solution is to only return a transaction id for each transaction request. The javascript can then store this transaction id. At the core of the javascript application would be an interval that would check the API for completed transactions and their corresponding responses. This approach is illustrated below.
Tuesday, August 11, 2009
Transactional Javascript API Interaction
Labels:
api
,
interaction
,
javascript
,
pattern
,
transaction
,
webapplication
Subscribe to:
Post Comments
(
Atom
)
Nice idea right there. With added response compression of some sort I think this will work very nice
ReplyDelete