SIP%20Torture%20Tests

1
SIP Torture Tests

• Robert Sparks
• dynamicsoft
• SIPPING IETF58

2
Significant Rewrite

• Removed tests based on deprecated 2543 features
  (such as absolute times in Expires)
• Added tests focusing on 3261
• Increased rigor on tests of escaping, non-ascii
  UTF-8, long or out-of-range values, and protocol
  extension points
• Added detail to discussion of each message
• SDP tests have been removed (to appear in an
  mmusic sdp-torture test draft)
• Conformed to ID-Nits

3
Reducing Ambiguity

• Introduced new markup
• allOneLine
• hex
• repeat
• Added bit-accurate archive
• Base64 encoded zip file of the real messages
• Filename indicated in the Message Details line
  of each example.

4
Created ReusableTools

• I-D encoded and bit-accurate messages are built
  from a common source.
• Content-Length is automatically calculated (when
  requested).
• Designed to integrate easily with xml2rfc.

5
Teasers

• ltallOneLinegt
• !interesting-Method0123456789_.indeed'
• sip1_unusual.URI(to-be!sure)isn'tit/cr
  azy?,/
• ithas1,weird!paswod_too.(doesn't-it)
• _at_example.com SIP/2.0
• lt/allOneLinegt
• Expires 1ltrepeat count100gt0lt/repeatgt
• Contact ltsipuser_at_host129.example.comgt
• expires280297596632815

6
Teasers

• ltallOneLinegt
• SIP/2.0 200 23 52
  lthexgtD0BDD0BE20D181D182
• D0BE20D0B4D0B5D0B2D18FD0BDD0BED181D182D0BE20
  D0B4
• D0B5D0B2D18FD182D18C202D20D0BFD180D0BED181D1
  82D0
• BED0B5lt/hexgt
• lt/allOneLinegt
• SIP/2.0 4294967301 better not break the
  receiver

7
Open Issues

• There are some production bugs
• Message filenames didnt make it into the draft
  as intended
• lthexgt expansions in the bit-accurate files arent
  accurate
• All messages are new or heavily edited and need
  careful review

8
Open Issues

• There are questions about specific messages in
  the open-issues section of the draft. These need
  list attention. One is reasonable to discuss
  here
• Are the long-value examples in 3.1.1.7 long
  enough?

9
Future Work

• Should the messages be normalized to support
  automation?
• Reusing common names for the initiator and
  recipient
• Making some identifier (Call-ID) unique to assist
  with debugging