DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG 
RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT
<http://issues.apache.org/bugzilla/show_bug.cgi?id=30273>.
ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND 
INSERTED IN THE BUG DATABASE.

http://issues.apache.org/bugzilla/show_bug.cgi?id=30273

The HTTP 1.1 processor ignores "Connection: close" on response





------- Additional Comments From [EMAIL PROTECTED]  2004-07-28 06:07 -------
While I will not reopen this bug, I simply want to put my $0.10 cents in here
(there's more than $0.02 cents of material here).

First, my opinion is that the servlet container simply cannot have the kind of
insight into the various needs of any given client/server system to do the
"Right Thing" (tm) with respect to the transport layer all of the time.  The
working defaults provided by the container may be good enough for the 80% of
typical work loads, however the other 20% will have to go elsewhere and/or deal
with the situation using cumbersome workarounds.  These kinds of limitations
designed into the servlet container will reduce its ability to become an upward
path for legacy systems and will in addition place artifical limits on systems
with unusual work loads.

Second, I would like to point out that your last argument sound suspiciously
like "this is too hard to do right, therefore I'll ignore the problem."  This
problem will not go away because you deem it difficult to do right by all cases
you can think of.  From our experience, most of the clients we are right now
having problems with appear to be Windows clients with a relatively recent
security patch installed.  In other words, this specific problem will likely
become a larger issue than it is now, and you will likely have to deal with
dropping in a fix anyway, because for *damn* sure M$ isn't going to care.

Third, I can appreciate that time constraints may prevent a full out fix for all
of the issues you have raised, however, many of these header issues you have
outlined are already not handled (at all) by the servlet container, and thus are
currently put into place by the servlet anyway.  In other words you are FORCING
servlets to deal with the transport layer *today*.

For example, the only way that I can push a gzipped response to the client
*right now* is to check to see if the client issued me an "Accept-Encoding"
header field containing "gzip" and then respond with a "Content-Encoding: gzip"
header field and gzipping the response.  This is ALL done at the servlet layer,
because the container isn't doing this for me, I had to write a subclass of
javax.servlet.http.HttpServlet to do the right thing for all of my servlets.

By your argument, the connector should have done this check by default for the
servlet (with an option to turn off this behavior, of course).

Because this situation exists today, if you (the connector programmers as a
group) decide to handle the transport level "Accept-Encoding" header field from
within the container layer, then you'll have to deal with the compatability
issues of servlets that already handle this behavior.

Thus, your "WONTFIX" is simply putting off the inevitable, because while I'm
probably in a small minority of programmers actually pushing "Connection: close"
to clients, I'm sure that I am not the only developer that had to write code for
gzipping response bodies.

-- Not to mention that you STILL leave other servlets who are expecting that the
container will manifest the HTTP protocol properly (even in the corner cases) as
being broken by (your) design.

This is a definite "SHOULDFIX" in my book.

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to