[Standards-JIG] What happened to the ACK proposal?

JD Conley jd.conley at coversant.net
Sat Aug 13 05:42:44 UTC 2005


> I noticed that under certain circumstances Java is not able to detect
that
> a
> connection was closed. In particular, this problem happens when a
> router/firewall/NAT is involved in the connection.

We see the same issue in .NET as well (which is one reason why we were
on the positive side of that proposal).  It often takes a _very_ long
time for the dead connection to time out, depending on the TCP settings
of the OS in question.  Luckily most endpoints do a reconnect after they
are disconnected so you generally know when to discard the previous
connection.

SoapBox deals with this using a configurable white space keep alive,
just like most clients do.  We only send it on the outgoing connection
as we assume that the other endpoint is managing their connection to us.
A single byte of XML white space is sent every 30 seconds by default
(it's a configurable setting) during the normal S2S status/cleanup
check.  This tends to keep connections through NAT alive and also notice
when things go awry fairly quickly -- which unfortunately does happen
quite often.

-JD Conley



More information about the Standards mailing list