At 11/6/02 1:14 PM, OpenSRS Replies wrote:

>On Monday November 4, 2002 it came to our attention that the
>'Transfers Away Notification' functionality in OpenSRS had not    
>been running correctly for four days - from October 31, 2002 until
>November 3, 2002.  At that time, we re-initiated the functionality
>and ran the notifications for requested transfers that had arrived
>as of October 31, 2002.

It unfortunately seems like this kind of thing has happened a few times 
recently. Perhaps some engineering resources could be dedicated to 
creating regression tests and monitoring systems that check that each 
system is working every so often with an end-to-end test.

I've found this tremendously useful in my business. For example, I have a 
fairly complicated Apache server setup, and I was always worried that I'd 
break something whenever I made a change to the httpd.conf file or 
upgraded Apache. So I setup a test account on the server for each type of 
client we have (a test account that works identically to a real account), 
then spent a day writing a program that makes several dozen different 
types of HTTP requests (every possibility I could think of -- CGI with 
POST, CGI with GET, PHP with POST, PHP with GET, SSL combinations of 
each, etc.) to each test account and compares the result to the 
"expected" result.

Now when I make a change to Apache, and at regular intervals otherwise, I 
run that program and I can guarantee that everything we support is 
working properly.

Same thing with mail: I have a cron job that injects mail to addresses at 
the test accounts every five minutes (via SMTP), and another that 
attempts to read the mail that was injected (via POP). If no mail can be 
read for ten minutes, an alarm is triggered and my pager goes off.

I'm a strong proponent of monitoring systems by their result (Did SMTP 
injected mail actually get delivered and read through the POP server? 
Does the HTML response contain the text that the PHP script should have 
generated?), rather than programs that check servers indirectly (Does the 
SMTP server answer on port 25? Does the HTTP server respond on port 80?).

In the case of transfer away notifications, OpenSRS could create a system 
that submits a transfer away request for a test domain every hour, and 
then reads the domain's admin mailbox and reseller's contact mailbox for 
the notification messages (and cancels the transfer request). Every hour, 
you should get one "transfer away" message and one "transfer has been 
canceled" message at each address: if any of those don't arrive for two 
hours, something's wrong. Similar things are possible for other parts of 
OpenSRS.

Anyway, just my 2 cents.


>The 'Transfers Away Notification' is now operating normally.  We
>would like to stress to you, that 'Transfers Away Notification' 
>should not be used as an alternative to 'Name Locking' for name
>security.  The 'Transfers Away Notification' functionality relies
>on too many sources of information outside of our control to be
>considered a security tool.  It is an informational tool only.
>In order to prevent unauthorized Transfers Away, we recommend that
>Resellers utilize the 'Name Locking' functionality in OpenSRS.

Has there been any progress on creating a message to resellers that will 
notify us when a domain has actually been transferred? This "Transfer 
Away Notification" that tells us that the process has been started is 
currently the only message we get, but it doesn't tell us if the transfer 
away was completed.

------------------------------------
Robert L Mathews, Tiger Technologies

Reply via email to