MR 1208

[James Browning]

On Fri, Feb 5, 2021, at 4:04 PM Hal Murray via devel <devel at ntpsec.org>
wrote:

> > What were you looking for in the branch?
>
> The code that would get the correct source address on server replies.
>

The short version is I looked and it is not there^d^d^d^d^d anywhere I can
see it.


> I don't even know if it is possible.  The current  code turns on the right
> option and then gets the source address when it reads a packet using
> recvmsg()
> or recvfrom().
>
> Is there a similar option to get the dest address?  Is there an option on
> send
> to specify the source address?
>

It would appear yes, at least on StackOverflow.[1]


> How much will that slow things down?  Would we be willing to add the extra
> sockets to get that much of a speedup?  How ugly will that code be
> compared to
> the extra sockets?
>

I don't know, I've been told to walk away.


> Can we rewrite the isc code that does all the work of keeping track of
> interfaces so that it is simpleer/smaller?
>

Almost certainly. I had a tin dog program that I copied that spits out a
list of interfaces. It'd need bulking out but I'm off.

> I tested it using ntpdig on a couple of machines running Kubuntu 20.04 and
> > one w/ macOS High Sierra. Nothing complicated though, I can't be
> bothered to
> > think of and set up cases.
>
> If you aren't willing to test your changes on interesting cases, then
> please don't work on this area.  Yes, the current code is ugly, but there
> is a reason for that.


 Okay.

[1] https://stackoverflow.com/a/5281706/9831509
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ntpsec.org/pipermail/devel/attachments/20210205/8166b3c6/attachment.htm>