[Bundy-hackers] libdns++ first items
神明達哉 jinmei at wide.ad.jpMon May 19 19:41:01 CEST 2014
- Previous message: [Bundy-hackers] libdns++ first items
- Next message: [Bundy-hackers] "production-ready" support for shared memory (mmap)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
At Mon, 19 May 2014 22:02:04 +0530, Mukund Sivaraman <muks at banu.com> wrote: > > Does this proposal separate repositories for libdns++ and the rest of > > bundy? If so, will the bundy repo maintain its own copy of > > exceptions, util, etc, or will it require the separated libdns++ to > > build? > > It will require the separated libdns++ to build as a > dependency. libdns++ will be a shared library used by Bundy, Kea, > queryperfpp, etc. libdns++ is a Bundy dependency anyway. I don't necessarily think so, since at least conceptually Bundy is a protocol independent infrastructure. If and when a ntp server implementation is added, requiring libdns++ will be unnecessary and redundant dependency. On the other hand, we could be realistic that Bundy will solely be a platform for DNS and DNS only (plus DHCP, possibly, if we decide to keep the BIND10-derived DHCP stuff in Bundy). With the reality that it'll be difficult to find development and maintenance resource even for the narrowed goal, such a specialization decision may make sense. But I'd like to make that decision first. A related thing: how can we avoid maintaining multiple copies of m4 macros for commonly used stuff, such as the availability of Boost, if we maintain two repositories? For these reasons/questions, I personally prefer keeping the single repository and just provide an option to build libdns++ only (or to allow completely excluding DNS related stuff). -- JINMEI, Tatuya
- Previous message: [Bundy-hackers] libdns++ first items
- Next message: [Bundy-hackers] "production-ready" support for shared memory (mmap)
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]