Actions:
|
2009-12-04 13:55 AEST by Arthur Barrett - It would be great if there was a simple way to test the client / server connection without the need to
have an account on the server - in particular to test:
* client to server tx time (small file) direct and via packetizing buffer
* server to client rx time (small file) direct and via packetizing buffer
* client to server tx time (larger file - bigger than MTU) direct and via packetizing buffer
* server to client rx time (larger file - bigger than MTU) direct and via packetizing buffer
* reverse dns test
* compare client time to server time (UTC)
* is reverse DNS enabled on server?
* is 'run as' set to something other than 'client user' on server?
* is 'domain' set to something other than 'default' on server?
I think that 'cvs info' is the 'best' place to do this - but would need to double / triple check that the
code doesn't pass through the reverse dns or auth code which could slow it down.
Ideally the client could report the actual times and then finish with a:
a) if the server IP address is not 'intranet' or 'local' (127.x, 10.x etc)
*** THESE TIMES ARE ACCEPTABLE|UNACCEPTABLE FOR A WAN ****
b) if the server IP address is 'intranet' (10.x etc)
*** THESE TIMES ARE ACCEPTABLE|UNACCEPTABLE FOR A INTRANET ****
c) if the server IP address is 'localhost' (127.x etc)
*** THESE TIMES ARE ACCEPTABLE|UNACCEPTABLE FOR A LOCALHOST ****
Obviously we'd have to make some sort of guess at the figures - figure a) would be the hardest.
If the client / server time is different by more than what Active Directory and Kerberos permit we could
provide a warning, and if the reverse DNS is enabled and the time taken is more than (some figure?)
then also advise that the admin should either add this client to the DNS or disable reverse DNS. |