Frequently Asked Questions

Q: ROCCO seems to support only ONE packet stream on the link over which it is used. Is that true? If so, why and what happens if more than one packet stream goes over the link?
A: Yes and NO!! The current ROCCO specification defines profiles to handle both single and various numbers of multiple packet streams. In the 00 and 01 drafts, only ONE profile was defined for each IP version and that was a profile without support for multiple flows. In the cellular environment, for which we are aiming, different packet streams are expected to be separated on different logical channels which means that stream separation is not needed as a part of the header compression scheme. Therefor we decided to first design compression profiles without support for multiple packet streams. However, now we have that support provided in various ways to always make the compression efficient.
Q: When the first ROCCO solution only supported ONE packet stream, was it really fair to compare its compression efficiency with CRTP?
A: For most cases YES, but maybe not for all. If you, as in our cellular environment, have packet stream separation already provided, it is fair to compare the two cases because with CRTP it is mandatory to have the CID field. If support for multiple packet streams is required, then the situation could be different. However, the profile defined in the first ROCCO specification was not designed for multiple packet streams and the modified profiles with support for a various number of streams do still have a smallest compressed header size of two octets. Another aspect when comparing the two schemes is that ROCCO requires less functionality from the link-layer, which means that with ROCCO it is possible to save some octets there.
Q: It is stated that ROCCO can handle several consecutive lost packets on the link without losing the context. What happens if too many consecutive packets are lost?
A: The basic principle is that when more consecutive packets than a ROCCO profile is designed to handle are lost on the link, then the context is updated on request just as for CRTP. However, this is supposed to be a rare case because the profile used over a certain link should be chosen to suit the loss characteristics of the link. On the link used in our evaluations, loss of more than three consecutive packets happened rarely. The current ROCCO profiles can handle at least 26 consecutive packet losses without losing context.