Interesting phenomenon with Windows 2003 SP1 RC1

Hi,

I'm just interested to here if anyone else has tried this odd configuration,
and if you got the same error. I don't need any help per se, but I found it
fascinating and quite annoying before realising a workaround.

Basically, I connected my test server to a LAN with a DHCP assigned address,
and a switch connected to the LAN was also connected to a DSL modem. This
way, I could establish a connection to our ISP from the test server if I
disabled the connection for the main server. BUT, it was only possible to
connect in the "user-style" Network connections way, not by Routing and
remote access. From netmon traces I found that it seemed the server never
responded to an IPCP configuration packet, but I could not find out why. I
was thinking for a while that SP1 had introduced a problem where RRAS was
not allowed to use the netman COM+ application, because error messages about
access denied that application also appeared in the System log at the same
time. I elevated the permissions for a while, but it didn't change anything.

Finally, I turned on full, debug-included logging of all RRAS. In
%windir%\tracing\rasipcp.log I found this gem:
[1976] 01-18 21:58:15:187: IPCP: Server address lookup...
[1976] 01-18 21:58:15:187: IPCP: RasSrvrQueryServerAddresses...
[1976] 01-18 21:58:16:203: IPCP: RasSrvrQueryServerAddresses done(52)
[1976] 01-18 21:58:16:203: IPCP: IpcpMakeConfigRequest
[1976] 01-18 21:58:16:203: IPCP: Deferred IpcpBegin error=52
[1976] 01-18 21:58:16:203: IPCP: IpcpProjectionNotification
[1976] 01-18 21:58:16:375: IPCP: IpcpEnd...

As a very similar configuration works fine on the main server, I wondered
what could cause this, if it was not a bug in the service pack. Well, did
you read above that the host adapter IP address (the one PPPoE was tunneled
on) was DHCP assigned? This is not true of the main server, as it *is* the
DHCP server. Said and done -- I reserved an IP at the main server and gave
the test server a static IP.

This single change, AFAIK, gave these results instead:
[1976] 01-18 22:11:59:296: IPCP: Server address lookup...
[1976] 01-18 22:11:59:296: IPCP: RasSrvrQueryServerAddresses...
[1976] 01-18 22:12:00:328: IPCP: RasSrvrQueryServerAddresses done(0)
[1976] 01-18 22:12:00:328:
IPCPns=0103a8c0,Wins=00000000,DnsB=00000000,Wins B=00000000,Server=0503a8c0,Mask=ffffffff
[1976] 01-18 22:12:00:328: IPCP: UI parameters...

And so it worked.

Does anyone know what error 52 is in this context? I've seen that it may
reference to duplicate network names, but I don't think that makes total
sense here, but on the other hand it doesn't seem completely irrelevant.

/Carl