9 Feb
2003
9 Feb
'03
9:36 p.m.
Folks,
There is a bug in ser, which results in different hashes to be
generated if some of the fields participating in hash generation has
extra spaces before value in reply that were not present in request,
i.e.:
request:
From: <vasia_pupkin(a)1.2.3.4>
reply:
From: <vasia_pupkin(a)1.2.3.4>
This leads to matching failure and as a result causes useless
retransmition, if tm module is used. For example, Vovida's SIP stack
is found to be guilty in inserting extra spaces into reply's headers
(which isn't really prohibited by the RFC).
Attached little patch should fix the problem. It would be nice to have
it integrated into the next release.
Thanks!
-Maxim
11:18 p.m.
It's already changed on CVS -- hdr->body points to body beginning,
whatever LWS are ahead of it.
-jiri
At 09:01 PM 2/9/2003, jh(a)lohi.eng.song.fi wrote:
Maxim Sobolev writes:
--
Jiri Kuthan http://iptel.org/~jiri/
Attached little patch should fix the problem. It
would be nice to have
it integrated into the next release.
i think it would be better to fix the parser so that it would delete
leading white space from headers.
-- juha
_______________________________________________
Serusers mailing list
serusers(a)lists.iptel.org
http://lists.iptel.org/mailman/listinfo/serusers
11:29 p.m.
At 08:36 PM 2/9/2003, Maxim Sobolev wrote:
Folks,
There is a bug in ser, which results in different hashes to be
generated if some of the fields participating in hash generation has
extra spaces before value in reply that were not present in request,
i.e.:
request:
From: <vasia_pupkin(a)1.2.3.4>
reply:
From: <vasia_pupkin(a)1.2.3.4>
This leads to matching failure and as a result causes useless
retransmition, if tm module is used.
I guess that if there is a reply-matching failure, it has some other
reason. Reply matching is based on value of topmost branch in Via and
does not depend on From. Hash calculation is only used for request
processing.
For example, Vovida's SIP stack
is found to be guilty in inserting extra spaces into reply's headers
(which isn't really prohibited by the RFC).
Attached little patch should fix the problem. It would be nice to have
it integrated into the next release.
Parser in the next release is changed to set hdr->body to its real beginning
and skip LWS. (on CVS now) Nevertheless, I guess that is not related to the
issue you observed.
-Jiri
7963
days inactive
7963
days old
3 comments
3 participants
participants (3)
-
jh＠lohi.eng.song.fi -
Jiri Kuthan -
Maxim Sobolev