[bind10-dev] Resolver Plans for 2013

[JINMEI Tatuya / 神明達哉]

At Fri, 7 Dec 2012 16:16:53 +0100,
Shane Kerr <shane at isc.org> wrote:

> I put a page on the developer Wiki with some ideas for developing a
> resolver in 2013:
> 
> http://bind10.isc.org/wiki/ResolverPlan2013
> 
> Please have a read, and send feedback.

A few random comments:

- I believe we should primarily focus on some core features.  There
  are already many good recursive server implementations, so no one
  will be interested in just another mediocre recursive server.  My
  original understanding is that the primary focus of this sub project
  was response performance.  That's fine, and in that case we should
  initially focus on performance, excluding/deferring some other
  features.
- In that sense, I don't think it a good approach to try to do the
  various thing listed in "Milestone 2".  If the primary focus is
  performance, I'd rather complete something like multi-tiered cache
  and lower priority for things like query tracing.
- Another thing we should consider from the beginning is
  "stand-aloneness".  We'll lose the market of "base component" of BSD
  variants if we keep the current architecture of BIND 10, which
  depends on various external libraries and requires Python run-time.
  One possible decision is to accept the consequence and declare we
  are only interested in being adopted as part of some package system.
  On the other hand, if we still want to keep this market, we'll at
  least need to make it possible that the resolver server can run in
  the "standalone" mode.  It wouldn't require Python (or msgq as a
  result of it).  It should also be able to be built with as minimum
  dependency; we shouldn't require Python, Botan, (probably)
  log4cplus, or sqlite3.  For this mode of build we'd probably include
  a copy of Boost.  If we choose this path, it affects the
  architecture at the fundamental level, so it should be taken into
  account from the initial design phase.

---
JINMEI, Tatuya