HttpRequest and timeout

[hugo.lafayette]

I use an HttpRequest in my web application to launch an heavy
computation server side. The result is then send back to the browser.
The problem I got is that when the computation took too much time, then
I never get the result. I have to reload the page to get it.

I assume it's because my socket connection has time out, but my event
listener does not seems to get any event. Maybe I miss something:

function xmlhttpChange()
{
// looking fo a xmlhttp "loaded"
for (var i in nodeId2xmlHttp) {
if ((nodeId2xmlHttp[i] != null) && nodeId2xmlHttp[i].readyState ==
4) {
if (nodeId2xmlHttp[i].status==200) {
nodeId2xmlHttp[i] = null;
} else {
alert("Problem retrieving XML data (nodeId: " + i + ")");
nodeId2xmlHttp[i] = null;
}
}
}
}

My question is : how to handle XMLHttpRequest timeout. Any pointers ?

[VK]

hu************@gmail.com wrote:

My question is : how to handle XMLHttpRequest timeout. Any pointers ?

Request timeout is a system setting (this is why it's called "system
timeout"). You cannot handle it in the standard security environment.
Moreover you even cannot guess reliably what timeout set for a
particular machine.

Very approximately you can be safe within 10sec interval. You can add
extra interval by using "something back" trick: return content header
right away (200, text/html etc.) then start your calculations and spit
out an empty string or someting every 5-10 sec into the open channel.

On a really long run you should use the method applied by online
booking systems (like Travelocity.com) Upon request you send back id of
the result file like sid2894838292, close connection, do calculations,
save results sid2894838292 and exit.

Page will try to read sid2894838292 with the instructed delay (10sec,
20sec, 30 sec)

Nothing reliable you can do with JavaScript only.

[hugo.lafayette]

Thanks for the quick and clear answer. It appears to me that http is
not a "connected" mode, and must not be use like if it was. So your
solution seems the way to go for me.

[VK]

hu************@gmail.com wrote:

Thanks for the quick and clear answer. You are very welcome.
It appears to me that http is
not a "connected" mode, and must not be use like if it was.

Very true. Just to have the whole picture: modern HTTP has Keep-Alive
request but it's support and implementation is something I wouldn't
suggest to trust at all. Especially in anyhow serious project and
definitely not in money transactions.