I am very tempted to use long wait http or periodic polling by the client to set up pseudo-sockets on the browser side, for an application that would be used publicly.
But then on the 2nd thought, I am thinking HTML 5 is here.
But on the 3rd thought, what is the percentage of browsers out there that remain non-HTML5 within 12 months, 24 months, 36 months? If there are at least 20% of browsers still incapable of HTML5, then I cannot depend on HTML5 because 20% of users not being able to access an application is a significant amount.
What do you think, how would your advice be (to me and to developers in general)?
Q1.
Is there any point in rigging in COMET into an application anymore? I am thinking of gwt comet - http://code.google.com/p/gwt-comet/.
Q2.
Should we release a new public application within the next 2 months that is dependent on HTML5 sockets and tell non-HTML5 browser users "sorry, your browser version cannot access this application"?
Or should we architect the apps to use communication like GWT RPC?
Q3.
I am also very distrustful of long wait http request. I have never used it before but I have a horrible feeling about it. I have been using 10 to 20 second client-side polling. Is long wait http request risky (risk of hanging a browser session)? Does long wait request present any additional security risk?