Attention is currently required from: pespin.
laforge has posted comments on this change by pespin. ( https://gerrit.osmocom.org/c/osmo-ggsn/+/38500?usp=email )
Change subject: doc: Document MTU features in User Manual and example config files ......................................................................
Patch Set 5:
(2 comments)
File doc/manuals/chapters/running.adoc:
https://gerrit.osmocom.org/c/osmo-ggsn/+/38500/comment/3dfa86e0_2df359db?usp... : PS5, Line 57: When running osmo-ggsn, the user must take network Maximum Transmission Unit : (MTU) into consideration, and configure it based on network specific setup. I think this is too strong a statement. There's nothing inherently wrong with IP fragmentation. It's just sub-optimal. People operating a 2G or 3G lab network probably never even remotely need to worry about that, as the bandwidth/pps is so low.
Yes, in general, IP fragmentation is usually avoided by network administrators (irrespective of GTP). But in some cases it may very well be a valid decision to say "no, we prefer to eat the overhead of the fragmentation if we can get an inner MTU of 1500 to our clients". For example, if you use some weird protocol or equipment that doesn't properly handle lower MTUs, or if you are worried about too many problems with path MTU blackholes, etc.
So IMHO the documentation should make it clear that all of this is an optimization problem, not a "I must get this right to make it work at all" kind of topic.
https://gerrit.osmocom.org/c/osmo-ggsn/+/38500/comment/1ef51759_6533398a?usp... : PS5, Line 122: OsmOGGSN * spelling (OsmO vs Osmo) * TCP MSS clamping of course only works for TCP. There are other protocols out there that may or may not have different mechanisms.
In general, it might be a good idea to also explicitly state that if it's possible to increase the outer link MTU to accomodate an inner tunnel MTU of 1500, it is much preferred over all this smaller inner MTU, MSS clamping, etc.