Chronological Thread 

< Chronological >      < Thread >

---

Even if your primary goal isn't performance you shouldn't totally sacrifice it
in favor of questionable API decisions.

The portability issues related to addresses are unavoidable (given that
the library uses them internally, even outside of adaptation.[ch])
but not unsolvable and should be dealt with properly. Using strings is a
costly over-simplification.

Marc

On Thu, Apr 05, 2001 at 11:24:36AM +0200, Michael Tüxen wrote:
> Dear all,
> we agreed on having only strings in the API. So we have to do the conversion
> for every UDP packet, but not for every SCTP Packet. If we do this also
> for every SCTP packet, well, then there is room for performance improvement.
> But keep in mind: This implementation is not optimized for performance at
> all.
> It was not a design goal.
> Best regards
> Michael
>
> --
> e-mail:
> Michael.Tuexen AT micmac.franken.de
>
> On Thu, Apr 5, 2001, Marc Boucher
> <marc+sctp AT mbsi.ca>
> wrote:
>
> >> > > - ain't it better to work with the sockunion struct instead of the
> >> > > char/byte array. At least the compiler get something more to do than
> >> > > generating assember code. It can warn you if you did
> >> > provide the wrong
> >> > > parameter to your function. (I like strong type checking anyway by a
> >> > > compiler)
> >> >
> >> > Well, we had some lengthy conversation with Wolfgang Schrüfer
> >> > and he felt it was better to have an sctp.h that did not include
> >> > any OS specific stuff (indeed it is plain C now).
> >> > Internally we use the struct sockunion - so there is one conversion
> >> > involved inside the library from string to struct sockunion
> >> > per new SCTP instance (plus another one in your application
> >> > maybe). As that should not happen very often I don't think it
> >> > is a performance problem.
> >
> >Really?
> >For each UDP and SCTP packet received, the sockunion is converted to a
> string,
> >(see adl_sockunion2str() calls in adaptation.c:dispatch_event())
> >which would have to be eventually reparsed by the callback.
> >This is unnecessary overhead.
> >
> >> I prefer having predefined structs.
> >
> >So do I. IMHO passing string-encoded addresses, especially in callbacks,
> >isn't a good idea.
> >
> >Marc
> >
> >> Need the converted IP address anyway caus it can be included in the
> SUA msg.
> >> Doesn't really matter to me,works (and change was confined to a single
> >> file
> >
> >> pffft...)
> >_______________________________________________
> >discussion mailing list
> >discussion AT sctp.de
> >http://www.sctp.de/mailman/listinfo/discussion
>
>
> _______________________________________________
> discussion mailing list
> discussion AT sctp.de
> http://www.sctp.de/mailman/listinfo/discussion
>

---

• [SCTP-Discussion] Is the timer_list initialised?, Coene Lode, 04/02/2001
  • <Possible follow-up(s)>
  • RE: [SCTP-Discussion] Is the timer_list initialised?, Coene Lode, 04/04/2001
    • Re: [SCTP-Discussion] Is the timer_list initialised?, Andreas Jungmaier, 04/04/2001
  • RE: [SCTP-Discussion] Is the timer_list initialised?, Coene Lode, 04/04/2001
    • Re: [SCTP-Discussion] Is the timer_list initialised?, Marc Boucher, 04/05/2001
      • Re(2): [SCTP-Discussion] Is the timer_list initialised?, Michael Tüxen, 04/05/2001
        • Re: Re(2): [SCTP-Discussion] Is the timer_list initialised?, Marc Boucher, 04/05/2001
      • Re: [SCTP-Discussion] Is the timer_list initialised?, Andreas Jungmaier, 04/05/2001
        • Re: [SCTP-Discussion] Is the timer_list initialised?, Marc Boucher, 04/05/2001
          • [SCTP-Discussion] Strings vs. Sockunion, Andreas Jungmaier, 04/05/2001
  • RE: [SCTP-Discussion] Is the timer_list initialised?, Wagner, Joel, 04/05/2001
    • Re: [SCTP-Discussion] Is the timer_list initialised?, Marc Boucher, 04/05/2001
  • RE: [SCTP-Discussion] Is the timer_list initialised?, Wagner, Joel, 04/05/2001

---

Archive powered by MHonArc 2.6.19+.