On Fri, Oct 5, 2012 at 12:22 AM, Howard Lewis Ship <[email protected]> wrote:


> That would not work in a cluster; different servers in the cluster

Yes I know but anyone who deploy to a cluster would be loudly advised
to set the directive to a known value, but for all the other
deployments there would be no reason to worry about.

> would not be able to read each other's streams.  Yes, usually, its
> based on sticky sessions, but even then, there's fail-over to
> consider.  Also, a server restart would not only lose client sessions,
> but would generate a new random key, so any forms would become
> unsubmittable even if they did not depend on server-side session
> state.

This is case is valid only for case where a user GET a form then
populate it while a new deploy is running and submit it after the
deploy is finished.

>> It could lowered the error to a warning too.
>
> To be honest, in a perfect world, it would blow up with an exception
> ... but that's too draconian, especially for something added this late
> in the game.  Warnings are completely ignored.

Remember we are in the case of a lazy developer who has not set the
value or even has not read the release notes nor the doc so.
I want to protect the lazy developer in the typical "one server with
all in" situation.

I've talked about lowering the error to warning in case of a
completely random generated value not for the current hardcoded on
which I agree that a RuntimeException would be more wise.

> I'm even tempted to add a message to the AlertManager service, to
> really get it in front of the developer's eyes, pronto.

If the current implementation is choosen I would prefer a
RuntimeException in any case.

Cheers
-- 
Massimo
http://meridio.blogspot.com

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to