Transport-Related Protocol-Design Issues - Lars Eggert
Transport-Related Protocol-Design Issues - Lars Eggert
Transport-Related Protocol-Design Issues - Lars Eggert
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
lide title<br />
40 pt<br />
subtitle<br />
24 pt<br />
Text<br />
24 pt<br />
level 2-5<br />
20 pt<br />
NAT and Middlebox Traversal (II)<br />
Another headache: middlebox state silently times out when<br />
a session has been idle for a while<br />
– There are no rules for how soon this happens<br />
Results in communication failures when an application<br />
resumes transmission<br />
Popular countermeasure: keep-alives<br />
– “send dummy traffic when there’s nothing real to send”<br />
– incredibly wasteful (esp. on battery operated devices)<br />
– isn’t guaranteed to keep the session alive, either<br />
– only sensible where failure is very, very bad (cf. FEC)<br />
Better countermeasure: robust session handling<br />
– Good protocols recover from communication failures anyway<br />
– failures due to middleboxes are like any other failure<br />
IETF-73, Minneapolis, MN, USA WG Chairs Training Lunch 16<br />
November 19, 2008<br />
Top right<br />
corner fo<br />
field-mar<br />
customer<br />
partner lo<br />
See Best<br />
for exam