[ITP] bind-9.6.0-1

[ITP] bind-9.6.0-1

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

This is quickly becoming a FAQ, so:

ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/bind-9.6.0-1-src.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/bind-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libbind9-devel/libbind9-devel-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libbind9-devel/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libbind9_50/libbind9_50-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libbind9_50/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libdns-devel/libdns-devel-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libdns-devel/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libdns50/libdns50-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libdns50/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisc-devel/libisc-devel-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisc-devel/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisc50/libisc50-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisc50/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccc-devel/libisccc-devel-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccc-devel/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccc50/libisccc50-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccc50/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccfg-devel/libisccfg-devel-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccfg-devel/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccfg50/libisccfg50-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/libisccfg50/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/liblwres-devel/liblwres-devel-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/liblwres-devel/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/liblwres50/liblwres50-9.6.0-1.tar.bz2
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/liblwres50/setup.hint
ftp://ftp.cygwinports.org/pub/cygwinports/release/bind/setup.hint


Yaakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAklzp6gACgkQpiWmPGlmQSPGtwCg5tclYcnXDQQx32zlAjFeHJpV
dCYAn2JXu04VcpkFz2FDGO9GMeZxVrR5
=6hR4
-----END PGP SIGNATURE-----

Re: [ITP] bind-9.6.0-1

[Christopher Faylor]

Woo hoo!

Thanks Yaakov!

cgf

Re: [ITP] bind-9.6.0-1

[Corinna Vinschen]

On Jan 18 22:54, Christopher Faylor wrote:
> 
> Woo hoo!
> 
> Thanks Yaakov!

Indeed.  Just go ahead, Yaakov.


Thanks,
Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: [ITP] bind-9.6.0-1

[Andrew Schulman]

A request:  I think a lot of people would find it useful if you were to split
out the client tools, in particular 'host', from the bind server package.  I
don't want to run a DNS server-- I just want to be able to use the 'host'
command.  Thanks, Andrew.

Re: [ITP] bind-9.6.0-1

[Corinna Vinschen]

On Jan 21 11:51, Andrew Schulman wrote:
> A request:  I think a lot of people would find it useful if you were to split
> out the client tools, in particular 'host', from the bind server package.  I
> don't want to run a DNS server-- I just want to be able to use the 'host'
> command.  Thanks, Andrew.

And here's another problem, Yaakov.

Why does the bind package require an /etc/resolv.conf file?  Isn't that
used only by the resolver lib?  And the resolver lib is part of Cygwin
1.7 anyway, or in the minires package for Cygwin 1.5.  Minires/Cygwin 1.7
don't need a resolv.conf file, so why does bind?!?


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: [ITP] bind-9.6.0-1

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Corinna Vinschen wrote:
> And here's another problem, Yaakov.
> 
> Why does the bind package require an /etc/resolv.conf file?  Isn't that
> used only by the resolver lib?  And the resolver lib is part of Cygwin
> 1.7 anyway, or in the minires package for Cygwin 1.5.  Minires/Cygwin 1.7
> don't need a resolv.conf file, so why does bind?!?

I was attempting to avoid a FAQ.  BIND is a DNS suite, and it needs to
know the address of a nameserver in order to do lookups.  The usage of
/etc/resolv.conf is well documented in both the manpages and the online
manual.


Yaakov


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAkl3WU4ACgkQpiWmPGlmQSNXyQCfXEFPRiLslEh0fwggqgZo7RKO
jNoAnAgTDGVHiR4Bgoavl5EEaYpO215X
=RCY6
-----END PGP SIGNATURE-----

Re: [ITP] bind-9.6.0-1

[Corinna Vinschen]

On Jan 21 11:20, Yaakov (Cygwin/X) wrote:
> Corinna Vinschen wrote:
> > Why does the bind package require an /etc/resolv.conf file?  Isn't that
> > used only by the resolver lib?  And the resolver lib is part of Cygwin
> > 1.7 anyway, or in the minires package for Cygwin 1.5.  Minires/Cygwin 1.7
> > don't need a resolv.conf file, so why does bind?!?
> 
> I was attempting to avoid a FAQ.  BIND is a DNS suite, and it needs to
> know the address of a nameserver in order to do lookups.  The usage of
> /etc/resolv.conf is well documented in both the manpages and the online
> manual.

Ok, but the Cygwin resolver lib also works without this file because the
nameserver information can be (and is) fetched from Windows.  Would it
be feasible to use the same method as the resolver lib so that we don't
have two libs using different methods?  The resolver lib uses
/etc/resolv.conf, but falls back to calling the Windows function
GetNetworkParam to fetch this information if the resolve.conf file
doesn't exist(*).


Corinna


(*) On NT4 the GetNetworkParam function is missing so the resolver
    checks the registry directly, but given the age of NT4 I wouldn't
    be too concerned...

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: [ITP] bind-9.6.0-1

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Corinna Vinschen wrote:
> Ok, but the Cygwin resolver lib also works without this file because the
> nameserver information can be (and is) fetched from Windows.  Would it
> be feasible to use the same method as the resolver lib so that we don't
> have two libs using different methods?  The resolver lib uses
> /etc/resolv.conf, but falls back to calling the Windows function
> GetNetworkParam to fetch this information if the resolve.conf file
> doesn't exist(*).

BIND does not use libresolv/minires, it uses its own code (in liblwres)
for parsing /etc/resolv.conf and doing the actual resolver operations.


Yaakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAkl3YicACgkQpiWmPGlmQSMTEQCdEHXZa8UFlHyVAKfioOMQt3w7
8X8AoJTjPNzTtCHUYEdZ9DzFHi3DDlCF
=Rmrf
-----END PGP SIGNATURE-----

Re: [ITP] bind-9.6.0-1

[Corinna Vinschen]

On Jan 21 11:58, Yaakov (Cygwin/X) wrote:
> Corinna Vinschen wrote:
> > Ok, but the Cygwin resolver lib also works without this file because the
> > nameserver information can be (and is) fetched from Windows.  Would it
> > be feasible to use the same method as the resolver lib so that we don't
> > have two libs using different methods?  The resolver lib uses
> > /etc/resolv.conf, but falls back to calling the Windows function
> > GetNetworkParam to fetch this information if the resolve.conf file
> > doesn't exist(*).
> 
> BIND does not use libresolv/minires, it uses its own code (in liblwres)
> for parsing /etc/resolv.conf and doing the actual resolver operations.

Well, yes.  That's why I'm asking the above question.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: [ITP] bind-9.6.0-1

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Corinna Vinschen wrote:
> Well, yes.  That's why I'm asking the above question.

I'm really not sure where you're going with this.  This isn't Cygwin
specific; BIND doesn't use libresolv on Linux either.


Yaakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAkl3b+wACgkQpiWmPGlmQSPqBwCgmiegh+It+YrLeBqrxRqQfupK
35wAn2oiQlA++BQFhXq1HZwNrAvn2FfI
=gBy1
-----END PGP SIGNATURE-----

Re: [ITP] bind-9.6.0-1

[Corinna Vinschen]

On Jan 21 12:56, Yaakov (Cygwin/X) wrote:
> Corinna Vinschen wrote:
> > Well, yes.  That's why I'm asking the above question.
> 
> I'm really not sure where you're going with this.  This isn't Cygwin
> specific; BIND doesn't use libresolv on Linux either.

I'm not asking to use the resolver lib.  I'm asking that the resolver
code in bind's libs uses the same method to fetch nameserver and
domain information as Cygwin's default resolver does.  Which is
basically

- use /etc/resolv.conf if available
- otherwise, use Windows' GetNetworkParam() function.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: [ITP] bind-9.6.0-1

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Corinna Vinschen wrote:
> I'm not asking to use the resolver lib.  I'm asking that the resolver
> code in bind's libs uses the same method to fetch nameserver and
> domain information as Cygwin's default resolver does.

You could just as easily make the same argument on Linux: glibc already
includes a resolver lib, so why shouldn't BIND use it?  I have yet to
see a Linux distribution patch bind to use libresolv, and I see little
difference between using libresolv and copying (part of) minires into BIND.

> Which is basically
> 
> - use /etc/resolv.conf if available
> - otherwise, use Windows' GetNetworkParam() function.

You asked originally if it was feasible to do so.  There is a Windows
version of BIND, so perhaps it is possible, but AFAICS accomodating both
would be difficult to maintain.  If we're going to expect every Cygwin
user to maintain an /etc/fstab, I really don't see why we can't expect
those few who want to use BIND to maintain an /etc/resolv.conf.


Yaakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAkl3q4oACgkQpiWmPGlmQSM3EQCeIaDh5+tE04p8Gq8ozWYsMEKn
7pwAoKXg4gPCQrRSsgMF4frUT1KDqQDA
=kDMr
-----END PGP SIGNATURE-----

Re: [ITP] bind-9.6.0-1

[Corinna Vinschen]

On Jan 21 17:11, Yaakov (Cygwin/X) wrote:
> Corinna Vinschen wrote:
> > I'm not asking to use the resolver lib.  I'm asking that the resolver
> > code in bind's libs uses the same method to fetch nameserver and
> > domain information as Cygwin's default resolver does.
> 
> You could just as easily make the same argument on Linux: glibc already
> includes a resolver lib, so why shouldn't BIND use it?  I have yet to
> see a Linux distribution patch bind to use libresolv, and I see little
> difference between using libresolv and copying (part of) minires into BIND.

Is my English really so bad?  I'm not (NOT) asking that bind uses the
minires/Cygwin resolver.  I'm just talking about the methods used to
fetch resolver information.

On Linux, the resolver in glibc and the resolver code in BIND are using
the same technique.  Which is, fetch the information from
/etc/resolv.conf.  That file is the only source of information on Linux
anyway.

On Cygwin, the resolver code in minires/Cygwin is right now using a
different technique than the resolver code in BIND.  Minires/Cygwin uses
the Win32 function GetNetworkParam() if /etc/resolv.conf does not exist.

The idea is that you enhance the BIND resolver code to do the same: Use
/etc/resolv.conf if it exists.  But if not, use GetNetworkParam() to
fetch the same information.  And that's a big difference since the user
doesn't have to duplicate this information, especially if it changes.
And using GetNetworkParam is really simple.  An #ifdef __CYGWIN__ with
about 30 lines of code.

> > Which is basically
> > 
> > - use /etc/resolv.conf if available
> > - otherwise, use Windows' GetNetworkParam() function.
> 
> You asked originally if it was feasible to do so.  There is a Windows
> version of BIND, so perhaps it is possible, but AFAICS accomodating both
> would be difficult to maintain.  If we're going to expect every Cygwin
> user to maintain an /etc/fstab, I really don't see why we can't expect
> those few who want to use BIND to maintain an /etc/resolv.conf.

Apples/Oranges.  POSIX mount points don't exist anywhere in Windows,
resolver information does.  Yes, /etc/passwd and /etc/group also contain
redundant information and I'm not happy with it.  But in contrast to
/etc/resolv.conf the information in this files also contains information
not available on the system by default.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: [ITP] bind-9.6.0-1

[Pierre A. Humblet]

At 12:37 PM 1/21/2009, Corinna Vinschen wrote:
 >On Jan 21 11:20, Yaakov (Cygwin/X) wrote:>
 > Corinna Vinschen wrote:
 >> > Why does the bind package require an /etc/resolv.conf file?  Isn't that
 >> > used only by the resolver lib?  And the resolver lib is part of Cygwin
 >> > 1.7 anyway, or in the minires package for Cygwin 1.5.  Minires/Cygwin 1.7
 >> > don't need a resolv.conf file, so why does bind?!?
 >>
 >> I was attempting to avoid a FAQ.  BIND is a DNS suite, and it needs to
 >> know the address of a nameserver in order to do lookups.  The usage of
 >> /etc/resolv.conf is well documented in both the manpages and the online
 >> manual.
 >
 >Ok, but the Cygwin resolver lib also works without this file because the
 >nameserver information can be (and is) fetched from Windows.  Would it
 >be feasible to use the same method as the resolver lib so that we don't
 >have two libs using different methods?  The resolver lib uses
 >/etc/resolv.conf, but falls back to calling the Windows function
 >GetNetworkParam to fetch this information if the resolve.conf file
 >doesn't exist(*).

There is another problem. If resolv.conf exists then minires and the cygwin 1.7
resolver default to using the name servers directly instead of using 
the internal
Windows resolver.
That's useful to those who don't want to use the default dns server 
of their isp, which
may offer a "DNS assistance" feature that maps non existent domains to a web
server operated by the ISP, while keeping dns assistance for their web browser.
http://log.psi.cc/2008/03/02/how-to-disable-verizon-dns-assistance/
http://netservices.verizon.net/portal/link/help/item?case=dns_assist_fios&partner=verizon&product=fios

If it's a problem we can change the behavior of the cygwin 1.7 
resolver to require
more than the existence of resolv.conf to enable the bypass.

Pierre

Re: [ITP] bind-9.6.0-1

[Corinna Vinschen]

On Jan 21 21:23, Pierre A. Humblet wrote:
> There is another problem. If resolv.conf exists then minires and the cygwin 1.7
> resolver default to using the name servers directly instead of using the 
> internal
> Windows resolver.
> That's useful to those who don't want to use the default dns server of 
> their isp, which
> may offer a "DNS assistance" feature that maps non existent domains to a web
> server operated by the ISP, while keeping dns assistance for their web browser.
> http://log.psi.cc/2008/03/02/how-to-disable-verizon-dns-assistance/
> http://netservices.verizon.net/portal/link/help/item?case=dns_assist_fios&partner=verizon&product=fios
>
> If it's a problem we can change the behavior of the cygwin 1.7 resolver to 
> require
> more than the existence of resolv.conf to enable the bypass.

I don't think so.  The above explained method sounds quite right to me.
Thas't what I expect if an /etc/resolv.conf file exists and I'm sure
that's what happens in BIND's resolver code as well.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: [ITP] bind-9.6.0-1

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Andrew Schulman wrote:
> A request:  I think a lot of people would find it useful if you were to split
> out the client tools, in particular 'host', from the bind server package.  I
> don't want to run a DNS server-- I just want to be able to use the 'host'
> command.  Thanks, Andrew.

I don't see that as a compelling argument to justify YA binary package.
 Just because you install a package doesn't mean you have to use every
single tool included therein.


Yaakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAkl3VqAACgkQpiWmPGlmQSPlxACg4KVpRwavGPK3A5nP1/B232o4
4vcAnjReO7mcydEBjK58A2p6e5ORwSj4
=iq7d
-----END PGP SIGNATURE-----

Re: [ITP] bind-9.6.0-1

[Andrew Schulman]

> > A request:  I think a lot of people would find it useful if you were to split
> > out the client tools, in particular 'host', from the bind server package.  I
> > don't want to run a DNS server-- I just want to be able to use the 'host'
> > command.  Thanks, Andrew.
> 
> I don't see that as a compelling argument to justify YA binary package.
>  Just because you install a package doesn't mean you have to use every
> single tool included therein.

At the risk of prolonging the argument, I'll just boldly claim without proof
that all but a small percentage of the people who install the bind package will
want only the client.  In that case, parsimony and security argue for not
installing the server on those users' hosts-- even if they choose not to run it.

Whether that argument is compelling enough to justify splitting the package is
of course your call.  In any case, thanks for packaging bind.

Andrew.