[Http-webtest-general] Re: Frames???

[Ilya M. <il...@ma...>]

>>>>> On Sun, 3 Nov 2002 11:21:10 -0800 (PST), "Dr. Tolga Yalcinkaya" <tya...@ya...> said:

TY> Ilya,
TY> Thanks for the response, albeit late, it is still
TY> better than none :-)

TY> The need for handling frames is due to single signon
TY> implementations.  Typically, one frame
TY> provides/maintains single signon session state and
TY> another frame provides application content and
TY> application session state.  In all cases, session
TY> states are maintained via hidden variables, cookies,
TY> javascripts etc.

TY> In my case, it is important that the test script
TY> invoke all frames for a given URL, in order to
TY> maintain session states in different frames.

TY> Although the method that you describe below does
TY> provide a simplistic approach to the problem, it does
TY> not address the fundamental problem of cookies and
TY> session variables in multiple frames for a given URL. 


TY> Don't you think it is more desirable to have a
TY> mechanism in place that deals with frames more
TY> elegantly, and perhaps seamlessly?  There might be
TY> particular reasons why a tester would like to handle
TY> frames manually.  Hence, it would be very desirable to
TY> have a parameter that turns frame support on and off.

TY> Any thoughts?

Hmm, how seamless do you want it to be? The problem is that
HTTP-WebTest is based on idea that one test corresponds to one
request/response pair. In short HTTP-WebTest is written on the HTTP
protocol level which knows nothing about frames. This is fundamental
design decision and it is not that easy to change. It is possible to
write HTTP-WebTest plugin that does automatic fetching of all frames
for each frameset url in test sequence but HTTP-WebTest architecture
will not allow you *easily* perform tests on these automatically
fetched frames (because now you have one test corresponding to many
request/response pairs).

Anyway if you only want session data from frames (i.e. cookies as I
understand) than probably it is ok. I'm not sure I have time to write
such plugin now but it should not be very hard and if you are
interested I can provide you with some directions.

P.S. And no, it seems that you cannot just replace LWP::UserAgent with
LWP::UserAgent::FramesReady in HTTP::WebTest and expect it work out of
box - as I understand they have incompatible APIs.

-- 
Ilya Martynov,  il...@ip...
CTO IPonWEB (UK) Ltd
Quality Perl Programming and Unix Support
UK managed @ offshore prices - http://www.iponweb.net
Personal website - http://martynov.org