bind dlz

[Walt Howard]

In article <bnspq8$nc1$1 at sf1.isc.org>, shaun bugler  <sb at hetzner.co.za> wrote:
>Does anyone have any comments on it?
>http://bind-dlz.sourceforge.net/
>I want to see what everyone else thinks of this before I go off and use
>it and find it sucks.
>
Well, the problems it claims to solve do not seem to be so intractable
as to require patching BIND.

1. BIND reads its input from text files.  BIND9 comes with a tool to
validate the files before using them.  Whatever else you think about
text files, there are a zillion tools for working on them.

2. BIND stores its data in memory.  Yes, it does, for fast access.
Rather than re-inventing a separate virtual-memory system for every
platform on which it runs, BIND uses the one that comes with the OS.
That's really the only portable way to do it.

3. It takes a while for BIND to parse all those input files at startup.
BIND9 uses a separate thread to do this, so it doesn't hurt response
time very much.

4. In a highly dynamic zone, complete reloads take too much time.
Quite aside from the fact that highly dynamic zones may indicate a
more fundamental system design problem, zones can be reloaded
individually.  I expect Kevin Darcy will weigh in with a comment
about doing all the updates with dynamic updates to named instead
of editing the zone files; he has big enough zones (that work well)
that such comments should be taken seriously.

So DLZ looks to be solving problems that might be better solved in
other ways.  But if you like their approach, give it a try, and
please let us know how you like it.


-- 
Walt Howard                         /"\  ASCII Ribbon Campaign
InterNet: whoward at ieee.org          \ /  No HTML in mail or news!
BellNet: +1 780 492 6306             X
                                    / \