You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hc.apache.org by Karl Wright <da...@gmail.com> on 2013/05/07 12:57:47 UTC
HTTPCLIENT-1350
Hi Oleg,
I know you were working on state cleanup for the 4.2.5 release. I think
HTTPCLIENT-1350 may be related to this. Would you be willing to have a
quick look, and recommend debug logging options so that we can get what you
think is needed to figure out the issue?
Thanks!
Karl
Re: HTTPCLIENT-1350
Posted by Oleg Kalnichevski <ol...@apache.org>.
On Tue, 2013-05-07 at 08:54 -0400, Karl Wright wrote:
> bq. The problem is that older (and broken HTTP/1.1) will have no chance to
> respond at all if the client waits forever for a 100 status.
>
> I understand this concern, but remember that the user had to explicitly
> enable USE_EXPECT_CONTINUE in the first place. If they were dealing with
> broken servers that did not properly support expect/continue, then why
> would they do that? In other words, we already have a good window into the
> user's intent here - no need to code for least common denominator.
>
> Karl
>
I remember that but still of opinion we should comply with the
requirements of the HTTP spec. Not every application can expect a
particular protocol compliance level from all potential target and proxy
servers. One might want to enable the 'expect: continue' handshake by
default and still expect the client to handle an occasional
non-compliant behavior.
Oleg
>
>
> On Tue, May 7, 2013 at 8:48 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
>
> > On Tue, 2013-05-07 at 08:38 -0400, Karl Wright wrote:
> > > If you are willing to wait forever on a socket, then waiting forever for
> > > the server to respond seems perfectly reasonable as well.
> > >
> > > It is clear from the w3c spec you quoted that expect/continue's primary
> > > mode is to *not* just continue. It allows that for compatibility with
> > > older implementations only.
> > >
> >
> > The problem is that older (and broken HTTP/1.1) will have no chance to
> > respond at all if the client waits forever for a 100 status. We should
> > follow the spec recommendation in this instance and have a finite
> > waiting time for a 100 status in my opinion.
> >
> > Oleg
> >
> > > Karl
> > >
> > >
> > >
> > > On Tue, May 7, 2013 at 8:35 AM, Oleg Kalnichevski <ol...@apache.org>
> > wrote:
> > >
> > > > On Tue, 2013-05-07 at 08:30 -0400, Karl Wright wrote:
> > > > > By default, I'd make it the same as the socket timeout. Makes
> > sense, no?
> > > > >
> > > >
> > > > Socket timeout is indefinite (zero) by default. We could pick the
> > > > greater value of socket timeout and, say, 5 seconds.
> > > >
> > > > Oleg
> > > >
> > > > > Karl
> > > > >
> > > > >
> > > > > On Tue, May 7, 2013 at 8:26 AM, Oleg Kalnichevski <ol...@apache.org>
> > > > wrote:
> > > > >
> > > > > > On Tue, 2013-05-07 at 08:03 -0400, Karl Wright wrote:
> > > > > > > Thanks - this allowed us to fix our problem, I believe...
> > > > > > > Only comment I have is that 2-3 seconds is perhaps too short as a
> > > > > > default.
> > > > > > > If someone enabled expect-continue, presumably they have a
> > reason to
> > > > use
> > > > > > > it, and such a short delay is inconsistent with that.
> > > > > > >
> > > > > > > Karl
> > > > > >
> > > > > > It is really hard to tell what default value would be more
> > appropriate.
> > > > > > 5 seconds?
> > > > > >
> > > > > > Oleg
> > > > > >
> > > > > > >
> > > > > > > On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <
> > olegk@apache.org>
> > > > > > wrote:
> > > > > > >
> > > > > > > > On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > > > > > > > > Hi Oleg,
> > > > > > > > >
> > > > > > > > > I know you were working on state cleanup for the 4.2.5
> > release.
> > > > I
> > > > > > think
> > > > > > > > > HTTPCLIENT-1350 may be related to this. Would you be
> > willing to
> > > > > > have a
> > > > > > > > > quick look, and recommend debug logging options so that we
> > can
> > > > get
> > > > > > what
> > > > > > > > you
> > > > > > > > > think is needed to figure out the issue?
> > > > > > > > >
> > > > > > > > > Thanks!
> > > > > > > > > Karl
> > > > > > > >
> > > > > > > > Karl
> > > > > > > > I do not think this is a bug. Please see my comment in the
> > JIRA.
> > > > > > > >
> > > > > > > > Oleg
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > ---------------------------------------------------------------------
> > > > > > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > > > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > > > > > >
> > > > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > ---------------------------------------------------------------------
> > > > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > > > >
> > > > > >
> > > >
> > > >
> > > >
> > > > ---------------------------------------------------------------------
> > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > >
> > > >
> >
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > For additional commands, e-mail: dev-help@hc.apache.org
> >
> >
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
Re: HTTPCLIENT-1350
Posted by Karl Wright <da...@gmail.com>.
bq. The problem is that older (and broken HTTP/1.1) will have no chance to
respond at all if the client waits forever for a 100 status.
I understand this concern, but remember that the user had to explicitly
enable USE_EXPECT_CONTINUE in the first place. If they were dealing with
broken servers that did not properly support expect/continue, then why
would they do that? In other words, we already have a good window into the
user's intent here - no need to code for least common denominator.
Karl
On Tue, May 7, 2013 at 8:48 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
> On Tue, 2013-05-07 at 08:38 -0400, Karl Wright wrote:
> > If you are willing to wait forever on a socket, then waiting forever for
> > the server to respond seems perfectly reasonable as well.
> >
> > It is clear from the w3c spec you quoted that expect/continue's primary
> > mode is to *not* just continue. It allows that for compatibility with
> > older implementations only.
> >
>
> The problem is that older (and broken HTTP/1.1) will have no chance to
> respond at all if the client waits forever for a 100 status. We should
> follow the spec recommendation in this instance and have a finite
> waiting time for a 100 status in my opinion.
>
> Oleg
>
> > Karl
> >
> >
> >
> > On Tue, May 7, 2013 at 8:35 AM, Oleg Kalnichevski <ol...@apache.org>
> wrote:
> >
> > > On Tue, 2013-05-07 at 08:30 -0400, Karl Wright wrote:
> > > > By default, I'd make it the same as the socket timeout. Makes
> sense, no?
> > > >
> > >
> > > Socket timeout is indefinite (zero) by default. We could pick the
> > > greater value of socket timeout and, say, 5 seconds.
> > >
> > > Oleg
> > >
> > > > Karl
> > > >
> > > >
> > > > On Tue, May 7, 2013 at 8:26 AM, Oleg Kalnichevski <ol...@apache.org>
> > > wrote:
> > > >
> > > > > On Tue, 2013-05-07 at 08:03 -0400, Karl Wright wrote:
> > > > > > Thanks - this allowed us to fix our problem, I believe...
> > > > > > Only comment I have is that 2-3 seconds is perhaps too short as a
> > > > > default.
> > > > > > If someone enabled expect-continue, presumably they have a
> reason to
> > > use
> > > > > > it, and such a short delay is inconsistent with that.
> > > > > >
> > > > > > Karl
> > > > >
> > > > > It is really hard to tell what default value would be more
> appropriate.
> > > > > 5 seconds?
> > > > >
> > > > > Oleg
> > > > >
> > > > > >
> > > > > > On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <
> olegk@apache.org>
> > > > > wrote:
> > > > > >
> > > > > > > On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > > > > > > > Hi Oleg,
> > > > > > > >
> > > > > > > > I know you were working on state cleanup for the 4.2.5
> release.
> > > I
> > > > > think
> > > > > > > > HTTPCLIENT-1350 may be related to this. Would you be
> willing to
> > > > > have a
> > > > > > > > quick look, and recommend debug logging options so that we
> can
> > > get
> > > > > what
> > > > > > > you
> > > > > > > > think is needed to figure out the issue?
> > > > > > > >
> > > > > > > > Thanks!
> > > > > > > > Karl
> > > > > > >
> > > > > > > Karl
> > > > > > > I do not think this is a bug. Please see my comment in the
> JIRA.
> > > > > > >
> > > > > > > Oleg
> > > > > > >
> > > > > > >
> > > > > > >
> > > ---------------------------------------------------------------------
> > > > > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > > > > >
> > > > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> ---------------------------------------------------------------------
> > > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > > >
> > > > >
> > >
> > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > For additional commands, e-mail: dev-help@hc.apache.org
> > >
> > >
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> For additional commands, e-mail: dev-help@hc.apache.org
>
>
Re: HTTPCLIENT-1350
Posted by Oleg Kalnichevski <ol...@apache.org>.
On Tue, 2013-05-07 at 08:38 -0400, Karl Wright wrote:
> If you are willing to wait forever on a socket, then waiting forever for
> the server to respond seems perfectly reasonable as well.
>
> It is clear from the w3c spec you quoted that expect/continue's primary
> mode is to *not* just continue. It allows that for compatibility with
> older implementations only.
>
The problem is that older (and broken HTTP/1.1) will have no chance to
respond at all if the client waits forever for a 100 status. We should
follow the spec recommendation in this instance and have a finite
waiting time for a 100 status in my opinion.
Oleg
> Karl
>
>
>
> On Tue, May 7, 2013 at 8:35 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
>
> > On Tue, 2013-05-07 at 08:30 -0400, Karl Wright wrote:
> > > By default, I'd make it the same as the socket timeout. Makes sense, no?
> > >
> >
> > Socket timeout is indefinite (zero) by default. We could pick the
> > greater value of socket timeout and, say, 5 seconds.
> >
> > Oleg
> >
> > > Karl
> > >
> > >
> > > On Tue, May 7, 2013 at 8:26 AM, Oleg Kalnichevski <ol...@apache.org>
> > wrote:
> > >
> > > > On Tue, 2013-05-07 at 08:03 -0400, Karl Wright wrote:
> > > > > Thanks - this allowed us to fix our problem, I believe...
> > > > > Only comment I have is that 2-3 seconds is perhaps too short as a
> > > > default.
> > > > > If someone enabled expect-continue, presumably they have a reason to
> > use
> > > > > it, and such a short delay is inconsistent with that.
> > > > >
> > > > > Karl
> > > >
> > > > It is really hard to tell what default value would be more appropriate.
> > > > 5 seconds?
> > > >
> > > > Oleg
> > > >
> > > > >
> > > > > On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <ol...@apache.org>
> > > > wrote:
> > > > >
> > > > > > On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > > > > > > Hi Oleg,
> > > > > > >
> > > > > > > I know you were working on state cleanup for the 4.2.5 release.
> > I
> > > > think
> > > > > > > HTTPCLIENT-1350 may be related to this. Would you be willing to
> > > > have a
> > > > > > > quick look, and recommend debug logging options so that we can
> > get
> > > > what
> > > > > > you
> > > > > > > think is needed to figure out the issue?
> > > > > > >
> > > > > > > Thanks!
> > > > > > > Karl
> > > > > >
> > > > > > Karl
> > > > > > I do not think this is a bug. Please see my comment in the JIRA.
> > > > > >
> > > > > > Oleg
> > > > > >
> > > > > >
> > > > > >
> > ---------------------------------------------------------------------
> > > > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > > > >
> > > > > >
> > > >
> > > >
> > > >
> > > > ---------------------------------------------------------------------
> > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > >
> > > >
> >
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > For additional commands, e-mail: dev-help@hc.apache.org
> >
> >
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
Re: HTTPCLIENT-1350
Posted by Karl Wright <da...@gmail.com>.
If you are willing to wait forever on a socket, then waiting forever for
the server to respond seems perfectly reasonable as well.
It is clear from the w3c spec you quoted that expect/continue's primary
mode is to *not* just continue. It allows that for compatibility with
older implementations only.
Karl
On Tue, May 7, 2013 at 8:35 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
> On Tue, 2013-05-07 at 08:30 -0400, Karl Wright wrote:
> > By default, I'd make it the same as the socket timeout. Makes sense, no?
> >
>
> Socket timeout is indefinite (zero) by default. We could pick the
> greater value of socket timeout and, say, 5 seconds.
>
> Oleg
>
> > Karl
> >
> >
> > On Tue, May 7, 2013 at 8:26 AM, Oleg Kalnichevski <ol...@apache.org>
> wrote:
> >
> > > On Tue, 2013-05-07 at 08:03 -0400, Karl Wright wrote:
> > > > Thanks - this allowed us to fix our problem, I believe...
> > > > Only comment I have is that 2-3 seconds is perhaps too short as a
> > > default.
> > > > If someone enabled expect-continue, presumably they have a reason to
> use
> > > > it, and such a short delay is inconsistent with that.
> > > >
> > > > Karl
> > >
> > > It is really hard to tell what default value would be more appropriate.
> > > 5 seconds?
> > >
> > > Oleg
> > >
> > > >
> > > > On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <ol...@apache.org>
> > > wrote:
> > > >
> > > > > On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > > > > > Hi Oleg,
> > > > > >
> > > > > > I know you were working on state cleanup for the 4.2.5 release.
> I
> > > think
> > > > > > HTTPCLIENT-1350 may be related to this. Would you be willing to
> > > have a
> > > > > > quick look, and recommend debug logging options so that we can
> get
> > > what
> > > > > you
> > > > > > think is needed to figure out the issue?
> > > > > >
> > > > > > Thanks!
> > > > > > Karl
> > > > >
> > > > > Karl
> > > > > I do not think this is a bug. Please see my comment in the JIRA.
> > > > >
> > > > > Oleg
> > > > >
> > > > >
> > > > >
> ---------------------------------------------------------------------
> > > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > > >
> > > > >
> > >
> > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > For additional commands, e-mail: dev-help@hc.apache.org
> > >
> > >
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> For additional commands, e-mail: dev-help@hc.apache.org
>
>
Re: HTTPCLIENT-1350
Posted by Oleg Kalnichevski <ol...@apache.org>.
On Tue, 2013-05-07 at 08:30 -0400, Karl Wright wrote:
> By default, I'd make it the same as the socket timeout. Makes sense, no?
>
Socket timeout is indefinite (zero) by default. We could pick the
greater value of socket timeout and, say, 5 seconds.
Oleg
> Karl
>
>
> On Tue, May 7, 2013 at 8:26 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
>
> > On Tue, 2013-05-07 at 08:03 -0400, Karl Wright wrote:
> > > Thanks - this allowed us to fix our problem, I believe...
> > > Only comment I have is that 2-3 seconds is perhaps too short as a
> > default.
> > > If someone enabled expect-continue, presumably they have a reason to use
> > > it, and such a short delay is inconsistent with that.
> > >
> > > Karl
> >
> > It is really hard to tell what default value would be more appropriate.
> > 5 seconds?
> >
> > Oleg
> >
> > >
> > > On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <ol...@apache.org>
> > wrote:
> > >
> > > > On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > > > > Hi Oleg,
> > > > >
> > > > > I know you were working on state cleanup for the 4.2.5 release. I
> > think
> > > > > HTTPCLIENT-1350 may be related to this. Would you be willing to
> > have a
> > > > > quick look, and recommend debug logging options so that we can get
> > what
> > > > you
> > > > > think is needed to figure out the issue?
> > > > >
> > > > > Thanks!
> > > > > Karl
> > > >
> > > > Karl
> > > > I do not think this is a bug. Please see my comment in the JIRA.
> > > >
> > > > Oleg
> > > >
> > > >
> > > > ---------------------------------------------------------------------
> > > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > > For additional commands, e-mail: dev-help@hc.apache.org
> > > >
> > > >
> >
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > For additional commands, e-mail: dev-help@hc.apache.org
> >
> >
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
Re: HTTPCLIENT-1350
Posted by Karl Wright <da...@gmail.com>.
By default, I'd make it the same as the socket timeout. Makes sense, no?
Karl
On Tue, May 7, 2013 at 8:26 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
> On Tue, 2013-05-07 at 08:03 -0400, Karl Wright wrote:
> > Thanks - this allowed us to fix our problem, I believe...
> > Only comment I have is that 2-3 seconds is perhaps too short as a
> default.
> > If someone enabled expect-continue, presumably they have a reason to use
> > it, and such a short delay is inconsistent with that.
> >
> > Karl
>
> It is really hard to tell what default value would be more appropriate.
> 5 seconds?
>
> Oleg
>
> >
> > On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <ol...@apache.org>
> wrote:
> >
> > > On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > > > Hi Oleg,
> > > >
> > > > I know you were working on state cleanup for the 4.2.5 release. I
> think
> > > > HTTPCLIENT-1350 may be related to this. Would you be willing to
> have a
> > > > quick look, and recommend debug logging options so that we can get
> what
> > > you
> > > > think is needed to figure out the issue?
> > > >
> > > > Thanks!
> > > > Karl
> > >
> > > Karl
> > > I do not think this is a bug. Please see my comment in the JIRA.
> > >
> > > Oleg
> > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > > For additional commands, e-mail: dev-help@hc.apache.org
> > >
> > >
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> For additional commands, e-mail: dev-help@hc.apache.org
>
>
Re: HTTPCLIENT-1350
Posted by Oleg Kalnichevski <ol...@apache.org>.
On Tue, 2013-05-07 at 08:03 -0400, Karl Wright wrote:
> Thanks - this allowed us to fix our problem, I believe...
> Only comment I have is that 2-3 seconds is perhaps too short as a default.
> If someone enabled expect-continue, presumably they have a reason to use
> it, and such a short delay is inconsistent with that.
>
> Karl
It is really hard to tell what default value would be more appropriate.
5 seconds?
Oleg
>
> On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
>
> > On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > > Hi Oleg,
> > >
> > > I know you were working on state cleanup for the 4.2.5 release. I think
> > > HTTPCLIENT-1350 may be related to this. Would you be willing to have a
> > > quick look, and recommend debug logging options so that we can get what
> > you
> > > think is needed to figure out the issue?
> > >
> > > Thanks!
> > > Karl
> >
> > Karl
> > I do not think this is a bug. Please see my comment in the JIRA.
> >
> > Oleg
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> > For additional commands, e-mail: dev-help@hc.apache.org
> >
> >
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org
Re: HTTPCLIENT-1350
Posted by Karl Wright <da...@gmail.com>.
Thanks - this allowed us to fix our problem, I believe...
Only comment I have is that 2-3 seconds is perhaps too short as a default.
If someone enabled expect-continue, presumably they have a reason to use
it, and such a short delay is inconsistent with that.
Karl
On Tue, May 7, 2013 at 7:25 AM, Oleg Kalnichevski <ol...@apache.org> wrote:
> On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> > Hi Oleg,
> >
> > I know you were working on state cleanup for the 4.2.5 release. I think
> > HTTPCLIENT-1350 may be related to this. Would you be willing to have a
> > quick look, and recommend debug logging options so that we can get what
> you
> > think is needed to figure out the issue?
> >
> > Thanks!
> > Karl
>
> Karl
> I do not think this is a bug. Please see my comment in the JIRA.
>
> Oleg
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
> For additional commands, e-mail: dev-help@hc.apache.org
>
>
Re: HTTPCLIENT-1350
Posted by Oleg Kalnichevski <ol...@apache.org>.
On Tue, 2013-05-07 at 06:57 -0400, Karl Wright wrote:
> Hi Oleg,
>
> I know you were working on state cleanup for the 4.2.5 release. I think
> HTTPCLIENT-1350 may be related to this. Would you be willing to have a
> quick look, and recommend debug logging options so that we can get what you
> think is needed to figure out the issue?
>
> Thanks!
> Karl
Karl
I do not think this is a bug. Please see my comment in the JIRA.
Oleg
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org