Content-Length header

[Cédric Damioli]

Hi,

I recently noticed that, at least for 2.1.x and 2.2.x, after any request
processing, Environment.commitResponse() is called which has the side
effect to compute the actual response body size and then set the
response content length.
While this is perfectly fine for GET requests, it's obviously useless
for OPTIONS and even wrong for HEAD requests.

Looking at code, an immediate workaround is to disable output buffering
but it's not satisfying.

Did someone encountered the same issue ?

I don't know exactly how to solve this without breaking legacy behaviour.
Any thoughts ?

Regards,

[Peter Hunsberger]

This sounds somewhat familiar, you may want to search the mailing list
archives to see if this has been discussed before.... Can commitResponse
tell what kind of request it is dealing with or if not can that be passed
in so that it knows?

Peter Hunsberger

Post by Cédric Damioli
Hi,
I recently noticed that, at least for 2.1.x and 2.2.x, after any request
processing, Environment.commitResponse() is called which has the side
effect to compute the actual response body size and then set the response
content length.
While this is perfectly fine for GET requests, it's obviously useless for
OPTIONS and even wrong for HEAD requests.
Looking at code, an immediate workaround is to disable output buffering
but it's not satisfying.
Did someone encountered the same issue ?
I don't know exactly how to solve this without breaking legacy behaviour.
Any thoughts ?
Regards,
--
Cédric Damioli
CMS - Java - Open Source
www.ametys.org

[Cédric Damioli]

I've searched archives but only found [1], which is only partly related.

Of course, commitResponse could be overridden in HttpEnvironment to
filter out some HTTP methods, but which ones ? Should we have a black
list, a white list, or a mean to configure it somewhere ? I don't want
to break some compatibility somewhere.
BTW, maybe only HEAD requests are concerned, after all ...

Cédric

[1]
http://cocoon.markmail.org/search/?q=content-length#query:content-length+page:4+mid:ddsqo5ezsstshh7a+state:results

Post by Peter Hunsberger
This sounds somewhat familiar, you may want to search the mailing list
archives to see if this has been discussed before.... Can
commitResponse tell what kind of request it is dealing with or if not
can that be passed in so that it knows?
Peter Hunsberger
Hi,
I recently noticed that, at least for 2.1.x and 2.2.x, after any
request processing, Environment.commitResponse() is called which
has the side effect to compute the actual response body size and
then set the response content length.
While this is perfectly fine for GET requests, it's obviously
useless for OPTIONS and even wrong for HEAD requests.
Looking at code, an immediate workaround is to disable output
buffering but it's not satisfying.
Did someone encountered the same issue ?
I don't know exactly how to solve this without breaking legacy behaviour.
Any thoughts ?
Regards,
--
Cédric Damioli
CMS - Java - Open Source
www.ametys.org <http://www.ametys.org>

[Peter Hunsberger]

Think you can likely just hard code it to skip HEAD and maybe OPTIONS if
you need to? Document the change and if we some day finally get a new
release out the door people will just have to test against their existing
applications. I don't see how they would break but maybe there are ways?

Peter Hunsberger

Post by Cédric Damioli
I've searched archives but only found [1], which is only partly related.
Of course, commitResponse could be overridden in HttpEnvironment to filter
out some HTTP methods, but which ones ? Should we have a black list, a
white list, or a mean to configure it somewhere ? I don't want to break
some compatibility somewhere.
BTW, maybe only HEAD requests are concerned, after all ...
Cédric
[1] http://cocoon.markmail.org/search/?q=content-length#
query:content-length+page:4+mid:ddsqo5ezsstshh7a+state:results
This sounds somewhat familiar, you may want to search the mailing list
archives to see if this has been discussed before.... Can commitResponse
tell what kind of request it is dealing with or if not can that be passed
in so that it knows?
Peter Hunsberger

Post by Cédric Damioli
Hi,
I recently noticed that, at least for 2.1.x and 2.2.x, after any request
processing, Environment.commitResponse() is called which has the side
effect to compute the actual response body size and then set the response
content length.
While this is perfectly fine for GET requests, it's obviously useless for
OPTIONS and even wrong for HEAD requests.
Looking at code, an immediate workaround is to disable output buffering
but it's not satisfying.
Did someone encountered the same issue ?
I don't know exactly how to solve this without breaking legacy behaviour.
Any thoughts ?
Regards,
--
Cédric Damioli
CMS - Java - Open Source
www.ametys.org

--
Cédric Damioli
CMS - Java - Open Sourcewww.ametys.org