Re: Path MTU discovery

Rick Jones <rick.jones2@xxxxxx> wrote:

If you can have something running on both sides, will the stacks of
interest update what they return in a getsockopt(TCP_MAXSEG) (might
have the macro wrong) call for a TCP connection?

Maybe, if I can stress out the local TCP sufficiently to make it send
maximum-size segments. I suppose I'm also reliant on the local host
actually doing path MTU discovery in its TCP (whereas messing about with
raw sockets will work regardless) but I guess all sane systems do this
nowadays.

Thanks for the suggestion.

-- [mdw]
.

Relevant Pages

  • Re: If I get two comcast internet subscriptions into my home, will I get double performance?
    ... >>but you won't get proper balancing. ... Perhaps you'd better take the time to understand how TCP works. ... you open a TCP connection from A, all packets for that connection must go ... Those boxes will not work for TCP traffic, ...
    (comp.dcom.modems.cable)
  • Re: Blocking receive and close
    ... TCP does this precisely to avoid ... the socket is destroyed immediately. ... The TCP connection hangs ... But just before you enter the recv function, ...
    (comp.programming.threads)
  • Re: Which server component?
    ... TCP connection. ... receive the messages and send a JSON answer string back to the client. ... My ruby tcp server is build up with this snippet of source code: ...
    (comp.lang.ruby)
  • Re: IP number question
    ... |>:I am talking about TCP. ... you can send responses with a range of sequence numbers, ... What is "The Ultimate Meme"? ...
    (comp.security.misc)
  • Re: iptables/backups/sound files fixed
    ... iptables has a timeout for idle TCP connections that was ... data transfers. ... increasing iptables' TCP timeout seems to be necessary ... Yes indeed, the data is transferred with one TCP connection, and the ...
    (Debian-User)