R
Richard Maher
Hi,
Can anyone point me to an RFC or any other spec, documentation, or standard
that describes what a HTTP browser/client should do when receiving an Applet
body (could be any file transmission for that matter) from a server.
My guess at what Internet Explorer 6 is doing is non-blocking socket reads
on the http connection and as soon as one returns an error (or zero bytes
transferred) then it clocks-it as a fail and tries again. If it fails N
times (Where 'N' appears to be 3) it starts asking for class files.
I only experience this problem when I set the transfer page size for the
server down to low values like 1K or 2K (and it can go up to 32K) but it
certainly appears that if the browser's Reads are quicker than the server
can get the stuff onto the wire, then it barfs.
I don't know if this is IE only (ie6 only) or all browsers, which is why I'm
asking for a pointer to standard or spec that says "On a http connection you
should not assume that a zero byte non-blocking read indicates a socket
close at the other end". Or some browser config parameter that can be set
like the Socket.connect(host, port, TIMEOUT)
Any thoughts, experiences, help?
Just increase the page size till it stops happening?
Cheers Richard Maher
Can anyone point me to an RFC or any other spec, documentation, or standard
that describes what a HTTP browser/client should do when receiving an Applet
body (could be any file transmission for that matter) from a server.
My guess at what Internet Explorer 6 is doing is non-blocking socket reads
on the http connection and as soon as one returns an error (or zero bytes
transferred) then it clocks-it as a fail and tries again. If it fails N
times (Where 'N' appears to be 3) it starts asking for class files.
I only experience this problem when I set the transfer page size for the
server down to low values like 1K or 2K (and it can go up to 32K) but it
certainly appears that if the browser's Reads are quicker than the server
can get the stuff onto the wire, then it barfs.
I don't know if this is IE only (ie6 only) or all browsers, which is why I'm
asking for a pointer to standard or spec that says "On a http connection you
should not assume that a zero byte non-blocking read indicates a socket
close at the other end". Or some browser config parameter that can be set
like the Socket.connect(host, port, TIMEOUT)
Any thoughts, experiences, help?
Just increase the page size till it stops happening?
Cheers Richard Maher