Future of the netgraph in DragonFly?

Matthew Dillon dillon at apollo.backplane.com
Sat Jun 2 10:29:44 PDT 2007


:Hello.
:
:As current developer and maintainer of the net/mpd4 (Multi-link PPP
:daemon for FreeBSD) project I would like to know DragonFly team ideas of
:the netgraph subsystem future. It looks frozen since it was branched
:from FreeBSD 4.x.
:
:I am not DragonFly user and currently have no interests there, but
:during long time there are many DragonFly support ifdefs exists in mpd
:sources. As FreeBSD 4.x is officially out of support it becomes annoying
:to be limited by it's requirements and I would like to remove that
:compatibility sometimes.
:
:Is the mpd project is interesting for DragonFly users? Should I keep
:hope that DragonFly will be updated to be in sync with FreeBSD netgraph
:at least in basic netgraph functionality or I will have to remove all
:DragonFly quirks from the code?
:
:The least thing I need this moment is the NG_NODELEN and other constants
:updating and somebody who could from time to time test new mpd version
:with DragonFly and maintain it ports/packages.
:
:-- 
:Alexander Motin

    Numerous projects have been removing conditionals incode for other
    projects for the same reason... having them makes the code unreadable
    and, strangely enough, less portable from a time perspective since even
    with the #ifdef's the code for OS's other then the one used by the
    project maintainer gets stale in fairly short order.

    Also, when projects do port code over they also tend to make an effort
    to normalize at least some of the APIs to reduce the complexity of
    doing the port rather then add in dozens of #if's.

    So my recommendation is that you remove the conditionals and focus on
    making the code as maintainable within FreeBSD as possible.

    If we wind up with a need to get a new multi-link ppp rev then we 
    will bring in the latest netgraph from FreeBSD as well.

					-Matt
					Matthew Dillon 
					<dillon at backplane.com>





More information about the Kernel mailing list