Metadata parser testing. Proxies & password

[Brent Putman]

On 3/31/14 9:43 AM, Cantor, Scott wrote:
> On 3/31/14, 5:12 AM, "Rod Widdowson" <rdw at steadingsoftware.com> wrote:
>> Do we care enough about this to build the requisite infrastructure.  My
>> instinct is that I am more nervous about having real user names and
>> passwords for our web services checked in (even if the associated account
>> is
>> locked right down) than I am about regressions and that the effort to
>> build
>> a complete test harness just to test these edge cases is not justified.
> I think we'd have to do it with embedded Jetty, probably, not with actual
> infrastructure. Doable, I'm sure, but probably not worth much time right
> now.

I have also been concerned about the lack of tests for real network
things like the HTTP basic auth and proxy settings.

I agree that the "right way" to do it (indeed for all unit testing in
general), is to not have them depend on things external to the project
(like web servers out in the world).  However, when I thought about
trying to do these tests, I realized that trying to do an embedded
Jetty, etc was beyond my knowledge at the time to do quickly.

To play devil's advocate: we already have basic HTTP metadata fetching
tests that hit an external web server (just plain anon HTTP requests to
the svn repo on shibboleth.net).   I imagine we could put some dummy
metadata files and dummy htpasswd-based username/password Basic auth on
the the Apache there, and be done in 10 minutes.  Figuring out how to do
it all in the project with Jetty I'm sure would take a lot longer
(although maybe Tom or someone else knows how to do this more quickly
than I).  So I guess it's really a question of how much effort and time
we want to put into doing it right, vs just getting it done.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://shibboleth.net/pipermail/dev/attachments/20140401/4466df88/attachment.html