Yes, this is exactly the sort of information I am looking for—thanks. I just did a pass through the code to refresh my memory, and the issue I have with it is that there's no architecture document, so you kind of have to figure that out by reverse engineering it, and there aren't a lot of comments. hncp_sd.c is actually pretty good—I don't think I'd looked at it before. E.g., one of the modules I looked at before I got this mail from you was pa_core.c. Actually it looks like there are some useful comments in the header files. But an introduction that explains how the code basically works would really help. Not the protocol—just what the concepts are and how they fit together. And debugging messages aimed at users would help as well—one of the easiest ways to learn how something works is to just run it with debugging maxed and see what it logs. But the log messages appear to be more about hncp than about what has been learned through hncp.
Anyway, I'll give it another look—you've already helped me quite a bit. On Fri, Nov 9, 2018 at 5:21 PM <[email protected]> wrote: > I guess what you want to do is publish and receive TLVs using DNCP, and > probably get a sense of the topology. > To do so, there mostly is a single header to look at: dncp.h > <https://github.com/sbyx/hnetd/blob/master/src/dncp.h> > > Here is a simple example of Wifi autoconfig we did for one of IETF’s > hackathon: https://github.com/Oryon/hnetd/blob/hackathon/src/hncp_wifi.c > As a more advanced example, I think the service discovery part will be > relevant to you: https://github.com/sbyx/hnetd/blob/master/src/hncp_sd.c > > If you want to add configuration parameters available in OpenWrt, I’d > recommend you look at platform-openwrt.c > <https://github.com/sbyx/hnetd/blob/master/src/platform-openwrt.c> > > - Pierre > > Le 9 nov. 2018 à 08:52, Ted Lemon <[email protected]> a écrit : > > Made out of LEGO bricks. What you’re describing doesn’t match my > experience last time I looked at it. I will look again. This is why I > asked: because I need help. This is the most information I’ve gotten about > it. I’m sorry my comments come across as saying the software is bad. That’s > not what i was trying to communicate. > > On Fri, Nov 9, 2018 at 4:47 PM Markus Stenberg <[email protected]> > wrote: > >> First off, hnetd was team effort - me, Pierre Pfister and Steven Barth. >> >> Secondly, I did not particularly want to promote hnetd but 'existing >> implementations are bad, boo hoo' argument gets old and I think e.g. >> https://github.com/jech/shncpd is also quite sufficient. I use even >> https://github.com/fingon/pysyma 'in production' even now (admittedly >> not HNCP bits of it, but DNCP + SHSP :->). >> >> On 09.11.2018, at 2.03, Ted Lemon <[email protected]> wrote: >> > The issue with the code (IIRC) is that it requires cmake to compile, >> for no obvious reason, and cmake is hard to get working, so e.g. building >> it on MacOS X is a major porting task. And it depends on libraries that I >> don't have. And there's no layering of the configuration system aspect of >> the architecture—it just goes and bashes on interfaces and stuff (this is >> my recollection—I haven't looked at it in a year or so). >> >> There's some distinct parts that have well defined interfaces (as far as >> those 'well defined' go in C anyway) - DNCP (dncp_*), /HNCP code (hncp_*), >> prefix assignment code (pa_*), and operating system (platform_*) are all >> pretty loosely coupled. >> >> I have compiled the DNCP(/partial HCNP) bits on OS X with relatively >> minor modifications for my hobby project for example. >> >> > These are not bad things that Markus did. Markus was doing what he >> needed to do to get it running on OpenWRT. But these are real issues, >> which are preventing me from using the code. For someone who is not an >> HNCP expert to hack on the code is difficult, and would require substantial >> work. I could do it, but I have other things I'm working on. If the >> code were more modular, I could experiment with it. This is the problem I >> was trying to express. >> >> I have used it also unmodified on plain Linux number of times. If you >> want to use it on something more esotreic, I could bet on OS X port being >> doable but as lots of HNCP value comes from interfacing with system >> servers, I do not see anyone writing platform interface for it. >> >> I am curious about what 'more modular' code looks like. Separate DLLs? >> Single function call API between modules? Possibly made out of Lego bricks? >> >> Cheers, >> >> -Markus > > _______________________________________________ > homenet mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/homenet > > >
_______________________________________________ homenet mailing list [email protected] https://www.ietf.org/mailman/listinfo/homenet
