(Illustration by Gaich Muramatsu)
On Mon, Sep 08, 2003 at 12:58:31PM +0900, Stephen J. Turnbull wrote: > >>>>> "Jan" == Jan Harkes <jaharkes_at_cs.cmu.edu> writes: > > Jan> You can possibly work around this by having the names of your > Jan> servers in /etc/hosts so that DNS lookup succeed even when > Jan> disconnected. > > Already thought of that, not good enough. Apparently an actual Coda > connection to the server is required. If you go "cfs disconnect; ls > /coda" before accessing /coda in any way, you get the dangling > symlink. "cfs resconnect; cfs checkservers; ls /coda" -> Bingo! There must be something else wrong because that pretty much always worked AFAIK. In fact I just tested it. I started venus, which wasn't running yet. cfs disconnect and ls /coda. It was stuck for a while until codacon showed that the various servers were unreachable and, $ ls /coda/coda.cs.cmu.edu ls: /coda/coda.cs.cmu.edu/hello.txt: Connection timed out ls: /coda/coda.cs.cmu.edu/platforms: Connection timed out ls: /coda/coda.cs.cmu.edu/tmp: Connection timed out ls: /coda/coda.cs.cmu.edu/project: Connection timed out ls: /coda/coda.cs.cmu.edu/backup: Connection timed out ls: /coda/coda.cs.cmu.edu/info: Connection timed out usr I only have /coda/coda.cs.cmu.edu/usr/jaharkes/... stuff cached, so the colorizing ls gets timeouts on the attributes of uncached objects. Now a lot of stuff related to how we deal with servers will need to be changed for IPv6 support anyways. But the 32-bit ipv4 addresses are currently everywhere so it won't be an easy fix. JanReceived on 2003-09-08 01:38:32