 |
Choose from over 20 free newsletters |  |
 |
Compare prices on computing products |
.gif){kind=spacer} |
Researcher at WWW10 proposes 'push and pull' protocol |
 Stephen Lawson, IDG
News Service\Hong Kong BureauMay 03, 2001, 06:41 | ||||||||||||
HONG KONG - Two new protocols proposed by a researcher at the 10th International World Wide Web Conference (WWW10) here Thursday would provide a way to use both "push" and "pull" techniques to deliver timely information over the Internet in a way that can grow with the number of customers. The PAP (Push And Pull) and POP (Push Or Pull) protocols could become standard ways for information providers to push updates to clients in some cases, and let clients pull those updates from a server in other cases, said Krithi Ramamritham, a professor at the Indian Institute of Technology (IIT) in Bombay, who presented the concept in a refereed paper at WWW10. Over its four days, the conference will feature 78 such papers, which have been selected from among 390 proposed for delivery at the prestigious annual conference. Presented for the first time at WWW10, the proposal could be picked up by the conference hosting body, the World Wide Web Consortium (W3C), or another standards body, or by a vendor of server or browser software, said Ramamritham in an interview following his presentation. Sending up-to-date information from a central server to PC or mobile-device clients has been touted as a key promise of Web-based customer services such as online stock trading. In addition, the capability to automatically deliver data among servers may be a central part of business-to-business Web services. Push technology today lets servers send updated information out to clients -- or other servers -- to inform users or kick off automated operations such as buying or selling stock. The consumer of the information can set out static rules for when updates should be sent. "Pull" is the way clients or servers that need new information take the initiative to ask for it from the server where it resides. Ramamritham and fellow researchers at IIT and at the University of Massachusetts at Amherst found push technology is very timely and reliable -- up until it has too many clients to take care of. Then, the work of continually checking if each client's trigger points have been reached can overwhelm the server and it starts to fall behind. On the other hand, making clients pull down information or check for updates at specific time intervals, such as every 15 minutes, can mean the client asks when it doesn't have to or gets the data later than it should, Ramamritham said. "We have to resort to a combination of push and pull," he said. PAP and POP would let servers check for updates for each client as long as they can, but then automatically have some clients switch over to pulling down the data. Customers that agree to get dropped off the push service might pay a lower monthly fee than full-time push customers. They could even define their own rules for when the push service would drop off. "It works out for you under certain circumstances but not all the time," Ramamritham explained. However, having the pull function available to take over could pay off for everyone, he added, by making the service more resilient. If a server or network connection fails, a push-only service can leave clients in the dark. Pull software at the client could be set to check in with the server if no updates have come for a certain period of time. Then, a customer could at least know the silence is a technical problem and look elsewhere for the information. "What is clear is we can achieve both resiliency and scalability by using the same mechanism," Ramamritham said. PAP and POP do basically the same thing, with the difference being that PAP has code for keeping track of the number of clients and their status. POP doesn't have to keep this "state" information because it is designed for servers and clients to switch between push and pull on their own, under certain circumstances. Ramamritham sees the two protocols as flexible tools that can be "tuned," similar to TCP (Transmission Control Protocol), which provides for reliable transmission of packets over an IP (Internet Protocol) network. Current pull capability built in to the HTTP (Hypertext Transfer Protocol) 1.1 standard doesn't offer as much flexibility, and push functions offered by vendors such as Netscape Communications Corp. and Bang Networks Inc. don't provide for a combination of push and pull, he said. "Once they find out that push doesn't scale... they could come here," Ramamritham said. WWW10 continues through Saturday.The W3C, based in Cambridge, Massachusetts, can be reached via the Web at http://www.w3.org//. | ||||||||||||
