Would there be a separate buffer per connection ? Otherwise a broken listener would still cause all the others to block...
Also with the current release would a slow listener cause all the other listeners to slow down because it is not clearing messages from the buffer fast enough ? Thanks Jonathan On 3/7/07, Hiram Chirino <[EMAIL PROTECTED]> wrote:
In the next major version (trunk branch) we have made some major strides in this area so that messages just go to disk on your senario instead of being buffered in memory. But all buffers (even ones on disks) have limits and will eventually be reached. So you have to detect dead clients anyways to avoid running out of disk space.
