[jitsi-dev] Re: Time lapse until connecting voice


#1

Hi,

please excuse me for insisting.

Am I the only one with this problem? It seems like a big show stopper to me and so I wonder why there is no resonance. Should I open an issue in Jira?

Conrad

···

----- Ursprüngliche Nachricht -----
Von: Conrad Beckert
Gesendet: 02.11.11 14:20 Uhr
An: dev@jitsi.java.net
Betreff: [jitsi-dev] Time lapse until connecting voice

Hi, today I tested Jitsi with the following configuration: - Jitsi build.3757 - Siemens Gigaset - both connected to local Asterisk server (extention 2004 respective 2002) Observed: - Calling 2002 from Jitsi, the phone rings - Jitsi sounds the mocked touch tones and then sounds the calling tone - I wait for two tones and pick up -> phone silent- no hear, no talk - Jitsi keeps "ringing" - after two or three seconds the connection gets through->speaking possible In a real world situation the called party whould have hung up long ago. What's wrong? We have had this before. Thank you for help. Conrad

Bitte antworten an/Please reply to/Por favor conteste a:
conrad_b@iname.com


#2

Hi Conrad,

Sorry for delay.

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I don't have issue. And I frequently receive calls from other Jitsis to this phone.

Can you please send us _full_ logs (including pcap), see http://jitsi.org/index.php/Documentation/FAQ#logs. Maybe your Asterisk server sends the INVITE 200 OK or ACK with little delay which prevent Jitsi to go "connected".

Regards,

···

--
Seb

Le 10/11/11 11:08, Conrad Beckert a écrit :

Hi,

please excuse me for insisting.

Am I the only one with this problem? It seems like a big show stopper to me and so I wonder why there is no resonance. Should I open an issue in Jira?

Conrad

----- Ursprüngliche Nachricht -----

Von: Conrad Beckert

Gesendet: 02.11.11 14:20 Uhr

An: dev@jitsi.java.net

Betreff: [jitsi-dev] Time lapse until connecting voice

Hi,

today I tested Jitsi with the following configuration:

- Jitsi build.3757
- Siemens Gigaset
- both connected to local Asterisk server (extention 2004 respective 2002)

Observed:
- Calling 2002 from Jitsi, the phone rings
- Jitsi sounds the mocked touch tones and then sounds the calling tone
- I wait for two tones and pick up -> phone silent- no hear, no talk
- Jitsi keeps "ringing"
- after two or three seconds the connection gets through->speaking possible

In a real world situation the called party whould have hung up long ago.

What's wrong? We have had this before.
Thank you for help.

Conrad

Bitte antworten an/Please reply to/Por favor conteste a:
conrad_b@iname.com


#3

Hi Sebastian,

Sorry for delay.

Never mind. Better late than never :slight_smile:

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I don't
have issue. And I frequently receive calls from other Jitsis to this
phone.

I'll send you the logs by private mail. For some weird reason the delay occurs on my laptop but not on the desktop. (same network) The delay is not related to the protocol - a direct call from one to the other machine on XMPP without the Asterisk in between shows the same result.

Thank you for the help.

Conrad


#4

На 10.11.11 12:28, Conrad Beckert написа:

Hi Sebastian,

Sorry for delay.

Never mind. Better late than never :slight_smile:

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I
don't have issue. And I frequently receive calls from other Jitsis
to this phone.

I'll send you the logs by private mail. For some weird reason the
delay occurs on my laptop but not on the desktop. (same network) The
delay is not related to the protocol - a direct call from one to the
other machine on XMPP without the Asterisk in between shows the same
result.

There may be a problem opening your device then ... although XMPP is a
different story since the ICE negotiation is likely to continue after
the callee picks up the phone.

Emil


#5

На 10.11.11 12:28, Conrad Beckert написа:

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I
don't have issue. And I frequently receive calls from other Jitsis
to this phone.

I'll send you the logs by private mail. For some weird reason the
delay occurs on my laptop but not on the desktop. (same network) The
delay is not related to the protocol - a direct call from one to the
other machine on XMPP without the Asterisk in between shows the same
result.

There may be a problem opening your device then ... although XMPP is a
different story since the ICE negotiation is likely to continue after
the callee picks up the phone.

I encounter the same issue in a different setup : I am using jitsi
(build 3768) on a laptop (Ubuntu 10.04). I encounter this delay/time
lapse problem every time I am called from regular phones via my SIP
provider (ippi). And also every time I call some regular phone line via
my SIP provider. And since these are "real life" calls, the other person
on the line sometimes hangs up because of the delay before they can hear
me. Or I am often asked "do you have a problem with your phone ? I have
been talking but you didn't seem to hear me ?". This is rather annoying.

We already mentioned this problem in this thread :

http://java.net/nonav/projects/jitsi/lists/dev/archive/2011-09/message/93

And Emil then said :

We'll have a look at the delay one of these days.

This problem does not seem to be referenced in the bug tracker.
Shouldn't we open a Jira ticket rather than keep mentioning it via email ?

Do you know of any work around ?

Cheers

···

Le 10/11/2011 12:43, Emil Ivov a écrit :

--
Jean Millerat
Wecena, rendre vos informaticiens fiers d'améliorer le monde
Jean.Millerat@wecena.com +33 1 77 455 344 Jean.Millerat@gmail.com


#6

I have experienced this issue myself. When debugging call setup I typically see a few java.lang.reflect.InvocationTargetException exceptions printed to the screen before an audio connection is established. I've narrowed in on two singleton objects (RTPManager & I believe DocumentBuilderFactory) that reference a function called newInstance(). Both object APIs have multiple implementations which they'll try to discover and invoke at runtime through reflection. The errors occur when these newInstance() functions attempt to invoke a non-existent package. There are multiple implementations of JMF, and RTPManager.newInstance(), as an example, will try to guess which ones you have installed, but it does this through reflection which makes it slow. This isn't a network issue, it seems to be more tied to how well your computer is performing and how quickly it could process through the code. My sense is that reflection is slow enough that on some machines this lapse is very noticeable, but its not so consistent. I hope that makes sense.

···

On Nov 28, 2011, at 8:27 AM, Jean Millerat wrote:

Le 10/11/2011 12:43, Emil Ivov a écrit :

На 10.11.11 12:28, Conrad Beckert написа:

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I
don't have issue. And I frequently receive calls from other Jitsis
to this phone.

I'll send you the logs by private mail. For some weird reason the
delay occurs on my laptop but not on the desktop. (same network) The
delay is not related to the protocol - a direct call from one to the
other machine on XMPP without the Asterisk in between shows the same
result.

There may be a problem opening your device then ... although XMPP is a
different story since the ICE negotiation is likely to continue after
the callee picks up the phone.

I encounter the same issue in a different setup : I am using jitsi
(build 3768) on a laptop (Ubuntu 10.04). I encounter this delay/time
lapse problem every time I am called from regular phones via my SIP
provider (ippi). And also every time I call some regular phone line via
my SIP provider. And since these are "real life" calls, the other person
on the line sometimes hangs up because of the delay before they can hear
me. Or I am often asked "do you have a problem with your phone ? I have
been talking but you didn't seem to hear me ?". This is rather annoying.

We already mentioned this problem in this thread :

http://java.net/nonav/projects/jitsi/lists/dev/archive/2011-09/message/93

And Emil then said :

We'll have a look at the delay one of these days.

This problem does not seem to be referenced in the bug tracker.
Shouldn't we open a Jira ticket rather than keep mentioning it via email ?

Do you know of any work around ?

Cheers

--
Jean Millerat
Wecena, rendre vos informaticiens fiers d'améliorer le monde
Jean.Millerat@wecena.com +33 1 77 455 344 Jean.Millerat@gmail.com


#7

Hey Jean,

На 28.11.11 14:27, Jean Millerat написа:

···

Le 10/11/2011 12:43, Emil Ivov a écrit :

На 10.11.11 12:28, Conrad Beckert написа:

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I
don't have issue. And I frequently receive calls from other Jitsis
to this phone.

I'll send you the logs by private mail. For some weird reason the
delay occurs on my laptop but not on the desktop. (same network) The
delay is not related to the protocol - a direct call from one to the
other machine on XMPP without the Asterisk in between shows the same
result.

There may be a problem opening your device then ... although XMPP is a
different story since the ICE negotiation is likely to continue after
the callee picks up the phone.

I encounter the same issue in a different setup : I am using jitsi
(build 3768) on a laptop (Ubuntu 10.04). I encounter this delay/time
lapse problem every time I am called from regular phones via my SIP
provider (ippi). And also every time I call some regular phone line via
my SIP provider. And since these are "real life" calls, the other person
on the line sometimes hangs up because of the delay before they can hear
me. Or I am often asked "do you have a problem with your phone ? I have
been talking but you didn't seem to hear me ?". This is rather annoying.

We already mentioned this problem in this thread :

http://java.net/nonav/projects/jitsi/lists/dev/archive/2011-09/message/93

And Emil then said :

We'll have a look at the delay one of these days.

This problem does not seem to be referenced in the bug tracker.
Shouldn't we open a Jira ticket rather than keep mentioning it via email ?

Yes please go ahead. Note however that adding it to the tracker doesn't
automatically mean an immediate solution so if anyone, who's having the
issue has some time to look into this then don't hesitate.

Cheers,
Emil


#8

Hey Oren,

На 28.11.11 15:52, Oren Forer написа:

I have experienced this issue myself. When debugging call setup I
typically see a
few java.lang.reflect.InvocationTargetException exceptions printed to
the screen before an audio connection is established. I've narrowed in
on two singleton objects (RTPManager & I believe DocumentBuilderFactory)
that reference a function called newInstance(). Both object APIs have
multiple implementations which they'll try to discover and invoke at
runtime through reflection. The errors occur when these newInstance()
functions attempt to invoke a non-existent package. There are multiple
implementations of JMF, and RTPManager.newInstance(), as an example,
will try to guess which ones you have installed, but it does this
through reflection which makes it slow. This isn't a network issue, it
seems to be more tied to how well your computer is performing and how
quickly it could process through the code. My sense is that reflection
is slow enough that on some machines this lapse is very noticeable, but
its not so consistent. I hope that makes sense.

The problem originally reported here was referring to intervals of 2-4
seconds in the beginning of calls. I am not convinced that reflection
could account for more than 200 ms or 300 ms.

As I mentioned a few mails back, my personal guess is that this is most
probably related to the code that opens the capture and playback devices.

Cheers,
Emil

···

On Nov 28, 2011, at 8:27 AM, Jean Millerat wrote:

Le 10/11/2011 12:43, Emil Ivov a écrit :

На 10.11.11 12:28, Conrad Beckert написа:

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I
don't have issue. And I frequently receive calls from other Jitsis
to this phone.

I'll send you the logs by private mail. For some weird reason the
delay occurs on my laptop but not on the desktop. (same network) The
delay is not related to the protocol - a direct call from one to the
other machine on XMPP without the Asterisk in between shows the same
result.

There may be a problem opening your device then ... although XMPP is a
different story since the ICE negotiation is likely to continue after
the callee picks up the phone.

I encounter the same issue in a different setup : I am using jitsi
(build 3768) on a laptop (Ubuntu 10.04). I encounter this delay/time
lapse problem every time I am called from regular phones via my SIP
provider (ippi). And also every time I call some regular phone line via
my SIP provider. And since these are "real life" calls, the other person
on the line sometimes hangs up because of the delay before they can hear
me. Or I am often asked "do you have a problem with your phone ? I have
been talking but you didn't seem to hear me ?". This is rather annoying.

We already mentioned this problem in this thread :

http://java.net/nonav/projects/jitsi/lists/dev/archive/2011-09/message/93

And Emil then said :

We'll have a look at the delay one of these days.

This problem does not seem to be referenced in the bug tracker.
Shouldn't we open a Jira ticket rather than keep mentioning it via email ?

Do you know of any work around ?

Cheers

--
Jean Millerat
Wecena, rendre vos informaticiens fiers d'améliorer le monde
Jean.Millerat@wecena.com +33 1 77 455 344 Jean.Millerat@gmail.com

--
Emil Ivov, Ph.D. 67000 Strasbourg,
Project Lead France
Jitsi
emcho@jitsi.org PHONE: +33.1.77.62.43.30
http://jitsi.org FAX: +33.1.77.62.47.31


#9

Hey Emil,

I think your estimate (200ms) is probably correct as I was looking over some of the additional debug statements that I applied to the codebase. I also came across the function ProcessorUtility.waitForState(Processor processor, int state). Do you think this might be where the bulk of the setup is taking place?

Thanks.

-Oren

···

On Nov 28, 2011, at 11:08 AM, Emil Ivov wrote:

Hey Oren,

На 28.11.11 15:52, Oren Forer написа:

I have experienced this issue myself. When debugging call setup I
typically see a
few java.lang.reflect.InvocationTargetException exceptions printed to
the screen before an audio connection is established. I've narrowed in
on two singleton objects (RTPManager & I believe DocumentBuilderFactory)
that reference a function called newInstance(). Both object APIs have
multiple implementations which they'll try to discover and invoke at
runtime through reflection. The errors occur when these newInstance()
functions attempt to invoke a non-existent package. There are multiple
implementations of JMF, and RTPManager.newInstance(), as an example,
will try to guess which ones you have installed, but it does this
through reflection which makes it slow. This isn't a network issue, it
seems to be more tied to how well your computer is performing and how
quickly it could process through the code. My sense is that reflection
is slow enough that on some machines this lapse is very noticeable, but
its not so consistent. I hope that makes sense.

The problem originally reported here was referring to intervals of 2-4
seconds in the beginning of calls. I am not convinced that reflection
could account for more than 200 ms or 300 ms.

As I mentioned a few mails back, my personal guess is that this is most
probably related to the code that opens the capture and playback devices.

Cheers,
Emil

On Nov 28, 2011, at 8:27 AM, Jean Millerat wrote:

Le 10/11/2011 12:43, Emil Ivov a écrit :

На 10.11.11 12:28, Conrad Beckert написа:

I personnally use a Siemens Gigaset S675 IP phone with Jitsi and I
don't have issue. And I frequently receive calls from other Jitsis
to this phone.

I'll send you the logs by private mail. For some weird reason the
delay occurs on my laptop but not on the desktop. (same network) The
delay is not related to the protocol - a direct call from one to the
other machine on XMPP without the Asterisk in between shows the same
result.

There may be a problem opening your device then ... although XMPP is a
different story since the ICE negotiation is likely to continue after
the callee picks up the phone.

I encounter the same issue in a different setup : I am using jitsi
(build 3768) on a laptop (Ubuntu 10.04). I encounter this delay/time
lapse problem every time I am called from regular phones via my SIP
provider (ippi). And also every time I call some regular phone line via
my SIP provider. And since these are "real life" calls, the other person
on the line sometimes hangs up because of the delay before they can hear
me. Or I am often asked "do you have a problem with your phone ? I have
been talking but you didn't seem to hear me ?". This is rather annoying.

We already mentioned this problem in this thread :

http://java.net/nonav/projects/jitsi/lists/dev/archive/2011-09/message/93

And Emil then said :

We'll have a look at the delay one of these days.

This problem does not seem to be referenced in the bug tracker.
Shouldn't we open a Jira ticket rather than keep mentioning it via email ?

Do you know of any work around ?

Cheers

--
Jean Millerat
Wecena, rendre vos informaticiens fiers d'améliorer le monde
Jean.Millerat@wecena.com +33 1 77 455 344 Jean.Millerat@gmail.com

--
Emil Ivov, Ph.D. 67000 Strasbourg,
Project Lead France
Jitsi
emcho@jitsi.org PHONE: +33.1.77.62.43.30
http://jitsi.org FAX: +33.1.77.62.47.31