Re: [homenet] How many people have installed the homenet code?

[Markus Stenberg <[e-mail]>]

On 26.4.2016, at 16.34, Rich Brown <[e-mail]> wrote:
> Ahhh... This is exactly the kind of advice I was looking for...
>> On Apr 26, 2016, at 9:04 AM, Markus Stenberg <[e-mail]> wrote:
>> 
>>> On 26.4.2016, at 15.09, Rich Brown <[e-mail]> wrote:
>>> 
>>> Thanks for this info. I have two primary interests here.
>>> 
>>> 1) I would like to stop farbling around with configuring subnets in my home. :-)
>>> 
>>> 2) With that knowledge in hand, I'll update the OpenWrt wiki, and include a procedure for getting the developer feed if necessary.
>>> 
>>> This all sounds pretty straightforward - and that the CC build is a good place to start. Thanks again!
>> Just as an advance warning, the CC version has some bugs;
> Are these bugs important? What problems would someone see? Or would those problems mostly affect people who're skilled in the project? As stated above, my goal is to have a set of clear instructions that “just work".

There are now 3 different hnetd versions in the wild, actually all slightly incompatible;

- (source-only) hnetd/master/HEAD: follows RFC7787/7788

- hnetd in openwrt-routing/master (DD): follows last draft-ietf-homenet-hncp-10

- hnetd in openwrt-routing/for-15-05 (CC): follows draft-ietf-homenet-hncp-08 

Someday when Steven has time we should probably bump both CC source and binary to hnetd/master/HEAD, following the RFC ports/TLV #s. 
DD I didn’t dare to bump as there seems to be some issue in OpenWrt trunk with the PCP firewall hole-punching (automated system tests with VMs fail as hole fails to appear; TBD why, but it happened to me last week).

Current version in openwrt-routing/master has minor bug or two; CC one has bunch of them, but despite those it probably mostly works. It does pass our automated system testcases, but there is one corner-case-ish crash and some endianness/IPv4 things that are not covered by them and that were fixed ‘later’ pre-current-DD-version.

> Plus, I realize I have a third goal:
> 
> 3) That people have a good experience the first time they try Homenet. (This makes using trunk/DD questionable - maybe the hnetd in trunk works great, but if they install a build of DD on a bad day, it’ll contribute to a bad experience, even though it's not hnetd's "fault".)

trunk/DD is definitely questionable with that goal. In general, I have mostly suggested just using CC for now.

> So my question for the considered wisdom of this group:
> 
> Is stock CC "good enough"? Or would CC+development version of hnetd be the best combination? If the latter, I will work toward those instructions. (And I may take Tore up on his offer to push a feed of the newer packages to Github…)

Probably the correct solution would be to just bump routing 15.05/master to hnetd/master/HEAD, as now the multicast address/port# should stay same and protocol also not change backward incompatibly going forward. While I can do it in the github repository (if I feel lucky enough that it works, see above note about PCP but I think it occurs only in DD which I do not care about that much), I am not sure how to get it to next CC binaries.

I am somewhat leery of pushing _anything_ until I have setup to test it out though, so this might wait until I get that Turris thingy :-)

Until that, I recommend just using CC and living with the bug or three or half dozen it has (git log is amusing reading :p).

Cheers,

-Markus