Zerbe John W <[EMAIL PROTECTED]> writes: > My own background is that of a mainframe assembly language programmer on a > team of people who supported a "home grown" TP monitor that is much like > and predates IBM's CICS product. By the way, this TP monitor is still in use > today and can still perform well on a machine that is running at 99% cpu busy. > I guess that makes me a bit sensitive to how the applications use/abuse > the supporting infrastructure. After having spent years preaching to > application programmers how queueing theory works and getting transaction > response times measured in the milliseconds, I still find it difficult to > imagine building an application where 10+ second response times are > "acceptable".
10+ second response times are not acceptable. That is the whole point of this thread! As you rightly intimate, asychronous coding is the way round such delays and that is what everyone in this thread has been talking about. Briefly the Java based suggestions to solve the problem have been: - dispatch a new Thread from the servlet request handler method to do the work and respond to the user with a "check status page" - variation on the above: respond to the user with a "check status page" but then continue using the servlet container's Thread to do the actual work - use JMS to queue the work to some other place - call a db stored proc to do the same (though this presumes your DB supports queuing... Oracle maybe?) -- Nic Ferrier http://www.tapsellferrier.co.uk ___________________________________________________________________________ To unsubscribe, send email to [EMAIL PROTECTED] and include in the body of the message "signoff SERVLET-INTEREST". Archives: http://archives.java.sun.com/archives/servlet-interest.html Resources: http://java.sun.com/products/servlet/external-resources.html LISTSERV Help: http://www.lsoft.com/manuals/user/user.html
