Shibboleth Developers

["Scott Cantor" <>]

> Ok. I've removed the boost requirement and implemented some KISS
> serialization.
> 
> That said, why don't you want to include this Storage Service in the
> distribution?

Well, frankly, the main reason was boost. I didn't expect you to give in so 
quickly. ;-)

I'm not against it, but I'm not about to go learn how to setup memcache and 
support people using it, and I sort of find that when I depend on things, I 
end up having to teach people how to use them.

> As I see it the only other way to cluster SPs is using the
> the odbc storage which doesn't even work well in unix+mysql.

In the majority of cases, clustering is an application issue. If an 
application isn't relying on the SP session, and most don't, what good is it 
to cluster that session by itself? It's usually easier to just use a 5 minute 
Shib session and rely on a few minutes of stickiness.

I just haven't seen more than one or two cases here where somebody was 
clustering *and* relying on the SP's session.
 
> Memcache is much easier to deploy and probably faster.

It's also not persistent, though, which is one of the reasons I did the ODBC 
thing. It's a slightly different use case.

Anyway, I don't care that much, I'm willing to check it in if people want it. 
(It would get done quicker if you have a tested autoconf patch to go with it.)

-- Scott