Firewall Wizards mailing list archives

Re:airgaps


From: Jonathan Squire <jsquirelists () crosswinds net>
Date: Sun, 22 Oct 2000 19:27:19 -0400

At 02:38 PM 10/19/00 -0400, you wrote:
> At what level is the "e-gap" unidirectional?  If it is truly unidirectional
> then how would I verify that the information was received correctly or even
> at all?  I should think that if I were to use it for storage of credit card
> transactions, it would behoove me to be able to guarantee that it was
> received/processed.  If I can't even do some level of confirmation across
> the gap how can I trust it except blind faith?  If I'm counting on that
> transaction for $$ I want to KNOW that it is saved safely.

When used in a truly unidirectional method... you can't get a transaction confirmation back... plain and simple.

There are some situations where it may be of some use though... to further on my previous example of receiving credit cards/personal information from a customer.

If we had the following scenario:
                                           |---------------------------------------|
V |
registration web server --> egap --> SQL Server               |
| | | filtered | V output |
External SQL server <--------------------- egap                       |
^ | | | V |
Order Web Server  -->----------------------------------------------------|

User input could be taken by the registration web server and forwarded to the internal SQL server... Things like their credit card number, address, phone number, etc... would be stored in here, a series of stored procedures could operate on this data on the inside and then push information that would be ok to have public out to the external SQL server (such as last 4 of the credit card and a confirmation that the needed information was sent in (address, telephone number, etc, etc) The order web server would then interact with the external database whenever it needed to display the information currently on file to the user, and when an order was placed it would send a transaction ticket to the internal sql server (or better yet an internal transaction server that would also verify the "application data" is correct), a confirmation would then be pumped out, the order server would just poll the external database, waiting for that confirmation, if it's not received in a time out, it would send a cancel transaction message inbound.

The above scenario better mimics the paper pushing method of placing an order... you go to a store, you pick out an item, you take it to the cash register, you present your credit card, they swipe your card, they present you with a receipt.



_______________________________________________
firewall-wizards mailing list
firewall-wizards () nfr com
http://www.nfr.com/mailman/listinfo/firewall-wizards


Current thread: