[Standards] s2s and gracelessly broken streams
fippo at goodadvice.pages.de
Thu Mar 29 08:07:04 UTC 2007
Justin Karneges wrote:
> Right. However, I think for this procedure we'd only want to probe against
> servers that we are no longer connected to, and only after some timeout. If
> the remote server is down, then the local server can return a presence error
> to the client. That one is easy. If the remote server is still up, but the
> remote contact has since gone offline, then the local server should have a
> timeout, whereby if no presence is received then the remote contact is
> assumed to be offline (I believe this is the only time a remote server would
> choose not to reply to a probe?).
> I'd propose, then, having 2 timeouts:
> 1) time to try probing (e.g. 10* minutes)
> 2) time to give up waiting for a reply, assume offline (e.g. 2* mins later)
> * Feel free to replace with better numbers.
Warning, heresy below:
I dont like the usage of the word 'assume'. What about making the reply
to a probe mandatory? Essentially I think it might be worthwile to
replace <presence/> based probing on s2s with iq-based ones (c2s does
not need to be changed for that).
This will result in increased traffic, yet - opposed to the 'probe
storm' - this is not unexpected traffic... if you send lots of probes
you know that there are going to be lots of answers. But you dont have
to make assumptions any more, as you will get replies (which are even
nicely tagged through the id attribute).
More information about the Standards