I had put code into the "Bilquist bridge" (bridge.c, version 2.8) to handle an IP address change. However, as this happened extremely late at night, I didn't want to wait for the sub-process to catch the change. So I whacked everything manually, dropping Bob a note, afterwards.

Since 2.8, I've done (I hope) a better job of strobing the information from the concurrently accessed bridge structure, which is now version 2.9. Both may be found for anonymous NFT on VENTI2::BRIDGE.C, if anybody wants to download. Johnny, I guess you may want to have a look when you get some time.

Be aware that VENTI2:: is running a much modified FAL, DAPLIB and GLXLIB, the current version being in internal alpha test, so let me know if you get what you feel are odd results.

On 1/17/23 10:22 AM, Paul Koning wrote:

I think I see why.  I looked at the code before and see that it is asking the name-resolution machinery to give it the current answer.  And that answer is supposed to be checked periodically.  But the particular method I used for the UDP cases doesn't do the timeout check.

I can fix that.

	paul

On Jan 17, 2023, at 10:14 AM, Robert Armstrong <bob@jfcl.com> wrote:

Paul Koning <paulkoning@comcast.net> wrote:
I think I covered all the various cases, including datagram based ones (GRE

and UDP).

If there are signs I missed something I'd like to get the details.

 "Bilquist bridge" circuits do not appear to recover from IP changes, even
when DNS is used.

Bob