22 Nov
2005
22 Nov
'05
4:39 a.m.
Imagine I have 50 salespeople, and I know the URIs of all of their
phones. When an INVITE comes in to SER, I want to keep trying to
forward it to the salespeople, one at a time (in series, not in
parallel), forever or until one of them accepts the call. (I
randomize the order.)
QUESTION 1:
At first, I tried rewriting the Request URI in a failure route each
time t_relay() returned failure, and then calling t_relay() again.
But t_relay() doesn't work that way -- when used in a failure route,
it requires you to append a branch, and apparently ignores any
changes in the Request URI after the failure happens.
OK, so in my failure route, I'll append a branch instead of rewriting
the Request URI. But the default value of MAX_BRANCHES in config.h
is only 12. I think this means I can never try to relay to more than
13 salespeople -- the first t_relay() call, plus 12 t_relay() calls
in failure routes. Is this correct?
QUESTION 2:
OK, so I'll make the value of MAX_BRANCHES greater than 12. But I
see in modules/tm/tm.c that MAX_BRANCHES cannot be given a value
greater than 31, since each branch occupies one bit in a 32-bit
bitmap. So how can I do 50 stateful relays? Or, for that matter, an
indefinite number?
Thank you for any help!
Tim
22 Nov
22 Nov
8:46 a.m.
Yes, you need to append a new branch to initiate a new transaction. Your
best bet, however, is to use lcr module and the q-value to serialize the
contacts. LCR is in CVS head and in the experimental module for 0.9.x.
g-)
----- Original Message -----
From: "Tim Klein" <tkpublic(a)timklein.fastmail.fm>
To: <serusers(a)lists.iptel.org>
Sent: Tuesday, November 22, 2005 2:39 AM
Subject: [Serusers] How to do a large number of relay attempts?
Imagine I have 50 salespeople, and I know the URIs of
all of their phones.
When an INVITE comes in to SER, I want to keep trying to forward it to the
salespeople, one at a time (in series, not in parallel), forever or until
one of them accepts the call. (I randomize the order.)
QUESTION 1:
At first, I tried rewriting the Request URI in a failure route each time
t_relay() returned failure, and then calling t_relay() again. But
t_relay() doesn't work that way -- when used in a failure route, it
requires you to append a branch, and apparently ignores any changes in the
Request URI after the failure happens.
OK, so in my failure route, I'll append a branch instead of rewriting the
Request URI. But the default value of MAX_BRANCHES in config.h is only
12. I think this means I can never try to relay to more than 13
salespeople -- the first t_relay() call, plus 12 t_relay() calls in
failure routes. Is this correct?
QUESTION 2:
OK, so I'll make the value of MAX_BRANCHES greater than 12. But I see in
modules/tm/tm.c that MAX_BRANCHES cannot be given a value greater than 31,
since each branch occupies one bit in a 32-bit bitmap. So how can I do 50
stateful relays? Or, for that matter, an indefinite number?
Thank you for any help!
Tim
_______________________________________________
Serusers mailing list
serusers(a)lists.iptel.org
http://lists.iptel.org/mailman/listinfo/serusers
9:07 p.m.
Thank you for the response, Greger! On your suggestion, I've read
about the LCR module, and it does seem useful here. But the LCR
functions (such as next_contacts()) also work by appending branches.
Am I correct in thinking that SER is not able to make more than 31
attempts to relay an INVITE, because it is not able to append more
than 31 branches? (This question may be better for the developers
mailing list, so I'm cross-posting there.)
If I am correct about that, then can anyone suggest a way to
statefully forward an INVITE without appending a branch?
Thank you in advance!
Tim
At 6:46 AM +0100 11/22/05, Greger V. Teigre wrote:
Yes, you need to append a new branch to initiate a new
transaction.
Your best bet, however, is to use lcr module and the q-value to
serialize the contacts. LCR is in CVS head and in the experimental
module for 0.9.x.
g-)
----- Original Message ----- From: "Tim Klein"
<tkpublic(a)timklein.fastmail.fm>
To: <serusers(a)lists.iptel.org>
Sent: Tuesday, November 22, 2005 2:39 AM
Subject: [Serusers] How to do a large number of relay attempts?
>Imagine I have 50 salespeople, and I know the URIs of all of their
>phones. When an INVITE comes in to SER, I want to keep trying to
>forward it to the salespeople, one at a time (in series, not in
>parallel), forever or until one of them accepts the call. (I
>randomize the order.)
>
>QUESTION 1:
>At first, I tried rewriting the Request URI in a failure route each
>time t_relay() returned failure, and then calling t_relay() again.
>But t_relay() doesn't work that way -- when used in a failure
>route, it requires you to append a branch, and apparently ignores
>any changes in the Request URI after the failure happens.
>
>OK, so in my failure route, I'll append a branch instead of
>rewriting the Request URI. But the default value of MAX_BRANCHES
>in config.h is only 12. I think this means I can never try to
>relay to more than 13 salespeople -- the first t_relay() call, plus
>12 t_relay() calls in failure routes. Is this correct?
>
>QUESTION 2:
>OK, so I'll make the value of MAX_BRANCHES greater than 12. But I
>see in modules/tm/tm.c that MAX_BRANCHES cannot be given a value
>greater than 31, since each branch occupies one bit in a 32-bit
>bitmap. So how can I do 50 stateful relays? Or, for that matter,
>an indefinite number?
>
>Thank you for any help!
>
>Tim
23 Nov
23 Nov
10:09 a.m.
Tim,
There are two different types of branches: parallel and serial. If you
branch in parallel, you append branches before t_relay and INVITEs are sent
out to ALL the branches at the same time. Serial branches are created in
on_failure and a new INVITE is sent when the first timed out, received busy
etc. AFAIK the limit to the number of branches is for parallel branches.
To your last question: A message is by definition (RFC) tightly connected to
a branch.
g-)
----- Original Message -----
From: "Tim Klein" <tkpublic(a)timklein.fastmail.fm>
To: "Greger V. Teigre" <greger(a)teigre.com>om>;
<serusers(a)lists.iptel.org>rg>;
<serdev(a)lists.iptel.org>
Sent: Tuesday, November 22, 2005 7:07 PM
Subject: Re: [Serusers] How to do a large number of relay attempts?
Thank you for the response, Greger! On your
suggestion, I've read about
the LCR module, and it does seem useful here. But the LCR functions (such
as next_contacts()) also work by appending branches.
Am I correct in thinking that SER is not able to make more than 31
attempts to relay an INVITE, because it is not able to append more than 31
branches? (This question may be better for the developers mailing list,
so I'm cross-posting there.)
If I am correct about that, then can anyone suggest a way to statefully
forward an INVITE without appending a branch?
Thank you in advance!
Tim
At 6:46 AM +0100 11/22/05, Greger V. Teigre wrote:
Yes, you need to append a new branch to initiate a
new transaction. Your
best bet, however, is to use lcr module and the q-value to serialize the
contacts. LCR is in CVS head and in the experimental module for 0.9.x.
g-)
----- Original Message ----- From: "Tim Klein"
<tkpublic(a)timklein.fastmail.fm>
To: <serusers(a)lists.iptel.org>
Sent: Tuesday, November 22, 2005 2:39 AM
Subject: [Serusers] How to do a large number of relay attempts?
>Imagine I have 50 salespeople, and I know the URIs of all of their
>phones. When an INVITE comes in to SER, I want to keep trying to forward
>it to the salespeople, one at a time (in series, not in parallel),
>forever or until one of them accepts the call. (I randomize the order.)
>
>QUESTION 1:
>At first, I tried rewriting the Request URI in a failure route each time
>t_relay() returned failure, and then calling t_relay() again. But
>t_relay() doesn't work that way -- when used in a failure route, it
>requires you to append a branch, and apparently ignores any changes in
>the Request URI after the failure happens.
>
>OK, so in my failure route, I'll append a branch instead of rewriting the
>Request URI. But the default value of MAX_BRANCHES in config.h is only
>12. I think this means I can never try to relay to more than 13
>salespeople -- the first t_relay() call, plus 12 t_relay() calls in
>failure routes. Is this correct?
>
>QUESTION 2:
>OK, so I'll make the value of MAX_BRANCHES greater than 12. But I see in
>modules/tm/tm.c that MAX_BRANCHES cannot be given a value greater than
>31, since each branch occupies one bit in a 32-bit bitmap. So how can I
>do 50 stateful relays? Or, for that matter, an indefinite number?
>
>Thank you for any help!
>
>Tim
4:12 p.m.
Tim Klein writes:
Imagine I have 50 salespeople, and I know the URIs of
all of their
phones. When an INVITE comes in to SER, I want to keep trying to
forward it to the salespeople, one at a time (in series, not in
parallel), forever or until one of them accepts the call. (I
randomize the order.)
i'm not sure if what you try to do is practical, since the calling phone
is not likely to wait so long for a final answer to invite.
-- juha
6926
days inactive
6927
days old
4 comments
3 participants
participants (3)
-
Greger V. Teigre -
Juha Heinanen -
Tim Klein