Eduardo Alvarenga wrote:
> Hi guys,
>
> I'm willing to port all DJB's public domain software, including djbdns,
> qmail, daemontools, etc.
> If made, It can be easily accepted on ports or, for historical reasons, i
> will not be imported?
I'm not a porter, I don't have final say in this at all, but:
1) DJBDNS is my DNS tool of choice.
2) I think BIND sucked before I learned BIND.
3) I think BIND sucks even more since I have learned it.
4) BIND sucks less since they learned a little from DJB
5) I highly respect DJB's "I'm doing it as I know is right,
I don't care what you think" attitude. ;)
So, I'm a fan of DJBDNS, but, I have to ask: Why?!?
Here are some of the issues that I know you will run into:
Do you "normalize" his directory layout, or do you maintain his
suggested layout (which has its merits, but is VERY non-hier(7)).
IF you normalize his layouts, you annoy the current DJBDNS
users and confuse the new ones. How do you propose to
announce this to the port user and orient them appropriately?
IF you don't normalize his directory layouts...I doubt they
will be imported when you slop three new directories in /.
Do you maintain the DJBDNS "dependency" on daemontools, even
though djbdns is one (er..two!) app(s) that just doesn't need
to be restarted when it fails (because it doesn't)?
IF you separate them, again, you will need to develop docs.
IF you leave them together...what value do you add to the
port?
This app was developed ON OpenBSD...it installs on OpenBSD
as well or better than any other OS. On the probably the
slowest machine you would be likely to want to run a DNS
server on (a P100 with a really slow disk attached to a really
slow IDE controler (wdc(4)!), it took ten minutes to build,
install and test. An experienced user might spend close to
that trying to figure out where things went, and a new user
won't find their life simplified.
LONG, LONG ago, probably around the 2.7 days, I used a DJBDNS
port on OpenBSD, and seriously regretted it. It was SO much
easier to simply install following DJB's instructions than it
was to find where things were put by the well-meaning porter.
I don't see how a port of the stock DJBDNS benefits anyone.
People who know and love DJBDNS won't use it, new users who
should learn it will be confused by it...so what is the point?
That being said, I do believe there are some people who are
working on taking the PD'ed DJBDNS package and bringing it
up-to-date. Assuming they normalize the directory layout and
document it appropriately and don't screw it up beyond all
recognition (those are three big "IFs" there) I'd have no
objection to that being made a port, and in fact, I'd welcome
it, but not a port just so we can say, "look! DJBDNS in
ports!".
I think similar arguments can be made against a qmail port.
Some of his utility programs (ucspi-tcp) would be really
handy ported, and I can not think of any issues involved
there. So there's a partial yes. :)
Nick.
