Re: [IBMiOSS] Node in-house auth?

That's sounds like a disaster waiting to happen.

-----Original Message-----
From: Mark Ford [mailto:Mark.Ford@xxxxxxxxxxx]
Sent: Tuesday, October 24, 2017 3:30 AM
To: IBMi Open Source Roundtable <opensource@xxxxxxxxxxxx>
Subject: Re: [IBMiOSS] Node in-house auth?

If the Kerberos routine that performs the authorisation returns you the user name (id@REALM) you can permit or deny actions in your application code via some sort of application authorisation/permissions file.

If you prefer to further authenticate the user has an actual IBM i ID you can use the eimGetTargetFromSource API to lookup the target ID from the source ID and verify it exists and is enabled.

If you actually want to perform the function under the authority of the IBM i user you could potentially perform a profile switch to the user, do the work, then switch back. Or in the case of database operations through the IBM API you might want to make the connection to the database as the user, issue the requests then disconnect. I should imagine there would be overhead with both of these approaches and there may be issues with async type operations and blocking. You wouldn't want the node process to start servicing a request for another user whilst still switched to the previous.

It makes you realise how much Apache is doing for you as it can take care of all of this by itself.

Disclaimer. I'm a sysadmin, not a developer, so these are just my thoughts/suggestions on an approach and these may or may not be practical.

Mark

nd think before you print this email.