Ed, experts
This feedback comes a tad late, but...
Are we really sure we want HttpFilter.doFilter(HttpSR, HttpSR) to be
protected?
Deep Java Enterprise stack traces end up at the end of quite a few jokes,
and adding a stack frame to potentially _every_ filter invocation will not
make the situation any better.
If we made the Http* doFilter method public, that could allow containers to
chop off the extra stack frame by calling the Http* methods directly, right?
Having suggested this feature, I would feel bad if it made Java stack
traces even harder to deal with than what they are today. So I'd love to
hear the feedback from the experts on this. Or at least just make sure my
worries are noted.
Cheers,
Eirik.
On Sep 29, 2015 12:45 AM, "Edward Burns" <edward.burns_at_oracle.com> wrote:
>>>>> On Thu, 24 Sep 2015 10:02:33 +0200, Eirik Bjørsnøs <eirbjo_at_gmail.com>
said:
Eirik> * Should HttpFilter::filter be abstract nor not? I kind of like
Eirik> the idea of having a no-op filter, but it also has benefits to
[...]
Eirik> least document in Javadoc that "There's almost no reason to
Eirik> override the filter(ServletRequest method.."), like HttpServlet
Eirik> documents for the service method.)
I've added that "there's no need to override this method. Also, I made
doFilter(HttpServletRequest req, HttpServletResponse resp) protected, to
match the way we do it on HttpServlet.
Eirik> * GenericFilter::getFilterInfo: I guess you added this to reflect
Eirik> Servlet::getServletInfo. But Filter has no getFilterInfo, so how
Eirik> much sense does it make to add it in GenericFilter? I suggest we
Eirik> skip it unless there are compelling reasons to include it.
I've dropped it.
Eirik> Otherwise, thank you for your quick turnaround on this!
You're too charitable. I hardly think nearly a year is quick
turnaround!
>>>>> On Thu, 24 Sep 2015 11:28:45 +0200, Eirik Bjørsnøs <eirbjo_at_gmail.com>
said:
EB> Ed,
EB> I noticed that HttpFilter.service(ServletRequest..) throws a
EB> ServletException if the request/response are not
EB> HttpServletRe[quest|sponse].
EB> Should HttpFilter.doFilter do the same?
EB> I don't know if / when this will ever happen. But I guess consistency
never
EB> hurts.
More than that. In specs, consistency is essential. I assume you mean
HttpServlet.service(ServletRequest...)? While it is true that the API
implementation does that, the javadoc doesn't say it does that. I think
that should be in the spec and I've created SERVLET_SPEC-143 for that.
I've committed it to the topic branch and also uploaded a new SNAPSHOT
[1].
Let me confer with Shing-wai and see if it's ready for the EG.
Thanks,
Ed
--
| edward.burns_at_oracle.com | office: +1 407 458 0017
| 28 Business days til JavaOne 2015
| 43 Business days til DOAG 2015
[1] https://maven.java.net/service/local/repositories/
snapshots/archive/javax/servlet/javax.servlet-api/4.0.
0-b01-SERVLET_SPEC-141-SNAPSHOT/javax.servlet-api-4.
0.0-b01-SERVLET_SPEC-141-20150928.223651-2-javadoc.jar/!/index.html