[httpclient] Constructing Cookies with null domains (again)

[Marc Saegesser <Ma...@apropos.com>]

There was some discussion last month about the expected results of
constructing a Cookie with a null domain.  I can't find the exact email
where the decision was reached, but I think the concensus was that we should
support a null domain and default to the request host.  There were a couple
test cases added to TestCookie.java to excersize constructing a cookie with
a null domain.  These tests are fail currently.

I'd like to revisit this issue and propose that we *don't* support
constructing cookies with a null domain.  In this case the constructor
should throw an IllegalArgumentException.  The reason is that the Cookie
constructor is not in a position to provide any reasonable default value for
domain because it does not have any idea what the request host is.

Marc Saegesser 

--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>

Re: [httpclient] Constructing Cookies with null domains (again)

[dIon Gillard <di...@multitask.com.au>]

Marc Saegesser wrote:

>There was some discussion last month about the expected results of
>constructing a Cookie with a null domain.  I can't find the exact email
>where the decision was reached, but I think the concensus was that we should
>support a null domain and default to the request host.  There were a couple
>test cases added to TestCookie.java to excersize constructing a cookie with
>a null domain.  These tests are fail currently.
>
There are three issues:
    - Cookie properties being null (domain, path etc)
    - Creating cookies using parse
    - Creating the cookie header using createCookieHeader

>I'd like to revisit this issue and propose that we *don't* support
>constructing cookies with a null domain.  In this case the constructor
>should throw an IllegalArgumentException.  The reason is that the Cookie
>constructor is not in a position to provide any reasonable default value for
>domain because it does not have any idea what the request host is.
>
But according to the cookie spec, domain is optional for the cookie 
header when sending to the client. So the Cookie class as a holder of 
the cookie before the request is sent, should allow null domains. But 
for parsing cookies from a header, or creating the header, I think you 
must specify a domain when calling these methods.

The tests are there to document current behaviour. I believe the tests 
fail due to nulls being passed on createCookieHeader or parse. These 
methods I think should be changed to throw NPEs if the params are null, 
and the tests changed accordingly.

>Marc Saegesser 
>
-- 
dIon Gillard, Multitask Consulting
http://adslgateway.multitask.com.au/developers




--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>