disable dnssec in bind resolver
Jan Buchholz
how i can disable dnssec in the bind resolver ? My firewall don´t let
packets with D0 flag through. I´ve tried 'dnssec-enable no;' , but
this don´t fix the problem.
Thanks,
Jan
Paul Wouters
> how i can disable dnssec in the bind resolver ? My firewall don´t let
> packets with D0 flag through. I´ve tried 'dnssec-enable no;' , but
> this don´t fix the problem.
I believe that only disables *serving* DNSSEC records.
I think you want 'dnssec-validation no;'
Paul
Jan Buchholz
sorry, 'dnssec-validation no;' is already configured, because that´s
the default.
Jan
Lightner, Jeff
Are you saying that "dnsec-validation no;" is in your named.conf or are you saying you don't believe it is necessary to set it there because by default validation is off? If the latter what does it hurt to try it? Obviously something isn't working the way you expect or you wouldn't have asked.
Jan
_______________________________________________
bind-users mailing list
bind-...@lists.isc.org
https://lists.isc.org/mailman/listinfo/bind-users
Proud partner. Susan G. Komen for the Cure.
Please consider our environment before printing this e-mail or attachments.
----------------------------------
CONFIDENTIALITY NOTICE: This e-mail may contain privileged or confidential information and is for the sole use of the intended recipient(s). If you are not the intended recipient, any disclosure, copying, distribution, or use of the contents of this information is prohibited and may be unlawful. If you have received this electronic transmission in error, please reply immediately to the sender that you have received the message in error, and delete it. Thank you.
----------------------------------
Jan Buchholz
my problem is, if a client want to resolve a ip-address from my
bind-server, the resolver set for some domains the D0 flag for the
question. And this behaviour don´t like my firewall.
Jan
2010/6/4 Lightner, Jeff <jlig...@water.com>:
Lightner, Jeff
Is it in your named.conf? That's a yes/no question.
Evan Hunt
> i mean the parameter is the default.
Actually, since 9.5.0, the default has been "dnssec-validation yes".
(Note, however, that DNSSEC validation doesn't occur unless the resolver
has a trust anchor configured. So you there has to be a "trusted-keys"
statement, a "managed-keys statement", or the "dnssec-lookaside auto"
option, or your resolver won't validate.)
Unfortunately, turning off validation won't help. A non-validating
recursive resolver still sets the DO bit--all that bit means is
"go ahead and send me DNSSEC data, it won't hurt me").
I'm pretty sure "dnssec-enable no" does suppress the DO bit. If it
doesn't, that's probably a bug.
If it doesn't, though, try "edns no". You can't have a DO bit if you
don't have a place to put one.
And, fix the broken firewall as soon as possible. :)
--
Evan Hunt -- ea...@isc.org
Internet Systems Consortium, Inc.
Paul Wouters
> I'm pretty sure "dnssec-enable no" does suppress the DO bit. If it
> doesn't, that's probably a bug.
Yeah, I thought the default changed when all those NAT routers proved buggy.
> If it doesn't, though, try "edns no". You can't have a DO bit if you
> don't have a place to put one.
This seems a bit like "my left leg hurts, so i stabbed my right leg".
> And, fix the broken firewall as soon as possible. :)
Now that is solid advise :)
Paul
Evan Hunt
> >don't have a place to put one.
>
> This seems a bit like "my left leg hurts, so i stabbed my right leg".
Exactly. Now you aren't lopsided.
R. Kevin Oberman
First, dns-validation is 'off' by default in all BIND versions. It's dnssec-enable that started defaulting to 'yes'.
Second, your firewall is simply broken. You will continue to have problems with DNS until you fix/replace it. I have not seen a recent firewall broken in this manner for a while, but this was quite common a couple of years ago.
For the moment, turning off dnssec-enable is probably your best hope, but it's not a fix and you are likeky to see continuing problems on a smaller scale until the firewall is fixed.
Sent from my Treo:
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
E. O. Lawrence Berkeley National Laboratory (LBNL)
obe...@es.net +1 510-486-8634
-----Original Message-----
From: Paul Wouters <pa...@xelerance.com>
Date: Friday, Jun 4, 2010 9:20 am
Subject: Re: disable dnssec in bind resolver
To: Evan Hunt <ea...@isc.org>
CC: bind-...@lists.isc.org
On Fri, 4 Jun 2010, Evan Hunt wrote:
> I'm pretty sure "dnssec-enable no" does suppress the DO bit. If it
doesn't, that's probably a bug.
Yeah, I thought the default changed when all those NAT routers proved buggy.
> If it doesn't, though, try "edns no". You can't have a DO bit if you
don't have a place to put one.
This seems a bit like "my left leg hurts, so i stabbed my right leg".
> And, fix the broken firewall as soon as possible. :)
Now that is solid advise :)
Paul
Alan Clegg
> First, dns-validation is 'off' by default in all BIND versions. It's
> dnssec-enable that started defaulting to 'yes'.
No, it isn't. The only reason that dnssec-validation appears "off" is
that without trust anchors, it doesn't do anything. Insert a trust
anchor and you validate, even without "dnssec-validation yes;" in your
configuration.
Really.
> Second, your firewall is simply broken. You will continue to have
> problems with DNS until you fix/replace it. I have not seen a recent
> firewall broken in this manner for a while, but this was quite common
> a couple of years ago.
100% agreed.
> For the moment, turning off dnssec-enable is probably your best hope,
> but it's not a fix and you are likeky to see continuing problems on a
> smaller scale until the firewall is fixed.
Yep.
AlanC
JINMEI Tatuya / 神明達哉
Jan Buchholz <96d...@googlemail.com> wrote:
> >> how i can disable dnssec in the bind resolver ? My firewall don´t let
> >> packets with D0 flag through. I´ve tried 'dnssec-enable no;' , but
> >> this don´t fix the problem.
> >
> > I believe that only disables *serving* DNSSEC records.
> >
> > I think you want 'dnssec-validation no;'
> sorry, 'dnssec-validation no;' is already configured, because that´s
> the default.
The DO bit is always set whenever the server includes an EDNS OPT RR
(I thought it was based on the specification, but don't remember which
sentence of which RFC says so).
So, your only choice is to completely disable EDNS:
server ::/0 {
edns no;
};
server 0.0.0.0/0 {
edns no;
};
As others said, however, I'd rather say "the fix is to upgrade/replace
the broken firewall". Please consider it only for a short term
workaround and seriously consider fixing the real problem.
---
JINMEI, Tatuya
Internet Systems Consortium, Inc.
Evan Hunt
> dnssec-enable that started defaulting to 'yes'.
Correct in the sense that there are no configured trust anchors, so
validation doesn't happen.
Incorrect in the sense that the "dnssec-validation" option *is* turned on
by default, from BIND 9.5.0 onward.
You're right that this isn't relevant to Jan's problem, though.
Doug Barton
> The DO bit is always set whenever the server includes an EDNS OPT RR
> (I thought it was based on the specification, but don't remember which
> sentence of which RFC says so).
Given that concern about whether or not it's a good idea to always send
DO=1 has come up in other contexts I for one would like to see chapter
and verse for why doing so is a MUST/SHOULD. If it turns out that DO=1
is not required I'd like to see a BIND option to turn it off.
Regarding the OP's situation, there are at least 2 problems. The first
being putting a firewall in front of a name server to start with, and
the second being that the firewall is broken. However I can think of
other reasons to want DO=0, especially in the age where having DNSSEC
records is going to be increasingly more common.
I have a guess at why ISC would want to enable it by default, and even
in the presence of an option to turn it off I'm still Ok with that
default. But if it's not a standards requirement to have it on, giving
the admin a choice would be a welcome thing.
FWIW,
Doug
--
... and that's just a little bit of history repeating.
-- Propellerheads
Improve the effectiveness of your Internet presence with
a domain name makeover! http://SupersetSolutions.com/
Paul Vixie
> I have a guess at why ISC would want to enable it by default, and even in
> the presence of an option to turn it off I'm still Ok with that default.
> But if it's not a standards requirement to have it on, giving the admin a
> choice would be a welcome thing.
this was, as you pointed out, a controversial decision. BIND implements the
"DO" bit as "this requestor will not vomit or crash if you include DNSSEC
metadata in the response". we believe that this supports the eventual goal
of near-universal DNSSEC deployment, in which it's foolish to treat "DO" as
"this requestor is explicitly interested in DNSSEC metadata on this answer".
the earlier we face the UDP fragmentation pain, the smaller that pain will
have been by the time we overcome it. same thing for validator bugs, zone
signing/resigning errors/expirations, and everything else that makes "always
set DO" seem unattractive today, to today's sysadmins, who aren't involved
in any DNSSEC deployment crusade and don't appreciate being co-opted for it.
unless a new IETF RFC comes along and disambiguates the meaning of "DO" such
that it's only to be set if the requestor thinks it has a reasonable shot at
validating the resulting metadata, i expect BIND to keep setting "DO" on all
EDNS requests it generates. and i don't think you can make a _public benefit_
argument that this is wrong even though there are _private benefit_ arguments.
--
Paul Vixie
KI6YSY
Doug Barton
> Doug Barton<do...@dougbarton.us> writes:
>
>> I have a guess at why ISC would want to enable it by default, and even in
>> the presence of an option to turn it off I'm still Ok with that default.
>> But if it's not a standards requirement to have it on, giving the admin a
>> choice would be a welcome thing.
>
> "DO" bit as "this requestor will not vomit or crash if you include DNSSEC
> metadata in the response". we believe that this supports the eventual goal
> of near-universal DNSSEC deployment, in which it's foolish to treat "DO" as
> "this requestor is explicitly interested in DNSSEC metadata on this answer".
>
> the earlier we face the UDP fragmentation pain, the smaller that pain will
> have been by the time we overcome it. same thing for validator bugs, zone
> signing/resigning errors/expirations, and everything else that makes "always
> set DO" seem unattractive today, to today's sysadmins, who aren't involved
> in any DNSSEC deployment crusade and don't appreciate being co-opted for it.
>
> unless a new IETF RFC comes along and disambiguates the meaning of "DO" such
> that it's only to be set if the requestor thinks it has a reasonable shot at
> validating the resulting metadata, i expect BIND to keep setting "DO" on all
> EDNS requests it generates. and i don't think you can make a _public benefit_
> argument that this is wrong even though there are _private benefit_ arguments.
Ok, so my guess as to ISC's motivations was pretty much on the mark, and
speaking with my "Guy who loves the Internet and wants to see things
work better for everybody" hat on, I am totally in agreement. That's why
I said I would have no problem with a theoretical DO knob defaulting to
"On."
With my business hat on though I can see at least 2 possible use cases
for DO=0. The first being related to this thread, "I can't/won't
fix/remove the firewall today, I just want my resolver to work." The
hapless user in that spot is either going to use another vendor, or go
back to the old version of BIND that "works." I know market share isn't
a _primary_ concern for BIND, but I would argue that the "go back to old
version" answer to this dilemma is something that we should all be
concerned about.
The other use case that leaps immediately to mind is "We do 42
scintillion DNS queries per second and our bandwidth cost has tripled in
the last 3 months! What in the name of J. Jonah Jameson is going on
around here?!?"
In all fairness, I don't have any actual clients telling me that DO=1 is
a problem for them, this is pure speculation on my part; although it's
speculation with a reasonable amount of experience behind it. In the
face of an actual client having actual DO=1 problems I would of course
encourage them to fix the underlying issue (and of course, to enable
DNSSEC). :) But if they can't/won't/etc ....
Doug (you kids with your newfangled contraptions, get off my lawn!)
Paul Vixie
> On 06/04/10 19:40, Paul Vixie wrote:
>> ...
>>
>> unless a new IETF RFC comes along and disambiguates the meaning of "DO"
>> such that it's only to be set if the requestor thinks it has a
>> reasonable shot at validating the resulting metadata, i expect BIND to
>> keep setting "DO" on all EDNS requests it generates. and i don't think
>> you can make a _public benefit_ argument that this is wrong even though
>> there are _private benefit_ arguments.
>
> ...
>
> With my business hat on though I can see at least 2 possible use cases for
> DO=0. The first being related to this thread, "I can't/won't fix/remove the
> firewall today, I just want my resolver to work."
it works. it's just slower because it has to fall back. this is one of the
reasons we fall back to BUFSIZE=512 before falling all the way back to DNS
(that is, turning EDNS off all together.)
> The hapless user in that spot is either going to use another vendor, or
> go back to the old version of BIND that "works." I know market share
> isn't a _primary_ concern for BIND, but I would argue that the "go back
> to old version" answer to this dilemma is something that we should all be
> concerned about.
that's been *very* rare on this point. ISC is concerned about relevance,
since we don't want to develop stuff that folks don't want to use. that's
*not* happening en masse in this case.
> ...
>
> In all fairness, I don't have any actual clients telling me that DO=1 is
> a problem for them, this is pure speculation on my part; ...
yes, i know that, because i'd see the other side of it if it was going on.
--
Paul Vixie
KI6YSY
Evan Hunt
> (I thought it was based on the specification, but don't remember which
> sentence of which RFC says so).
I was taken aback to read this, because I remembered seeing code in named
that clears the DO bit if "dnssec-enable" is "no":
if (!client->view->enablednssec) {
client->extflags &= ~DNS_MESSAGEEXTFLAG_DO;
[...]
}
Looking further, though, I see that Jinmei is correct. The above code
clears the DO bit in replies sent from an authoritative name server; it
doesn't apply to queries being sent by a resolver. Resolvers do indeed
set the DO bit unconditionally. Sorry for any confusion caused by my
earlier statement to the contrary.
Mark Andrews
In message <4C09C562...@dougbarton.us>, Doug Barton writes:
>
> Ok, so my guess as to ISC's motivations was pretty much on the mark, and
> speaking with my "Guy who loves the Internet and wants to see things
> work better for everybody" hat on, I am totally in agreement. That's why
> I said I would have no problem with a theoretical DO knob defaulting to
> "On."
>
> for DO=0. The first being related to this thread, "I can't/won't
> back to the old version of BIND that "works." I know market share isn't
> a _primary_ concern for BIND, but I would argue that the "go back to old
> version" answer to this dilemma is something that we should all be
> concerned about.
The resolver works. It figures out that it can't make the new style
queries and falls back to the old style queries. If the user is really
worried they can turn off EDNS and with that DO.
> The other use case that leaps immediately to mind is "We do 42
> scintillion DNS queries per second and our bandwidth cost has tripled in
> the last 3 months! What in the name of J. Jonah Jameson is going on
> around here?!?"
It's still a handful of zones that are signed. Referrals are a hundred or
so bytes bigger. Many still fit in 512 bytes. If your business isn't DNS
then you really won't notice. If it is DNS then you should be using that
data to verify the answers you are receiving.
> In all fairness, I don't have any actual clients telling me that DO=1 is
> a problem for them, this is pure speculation on my part; although it's
> speculation with a reasonable amount of experience behind it. In the
> face of an actual client having actual DO=1 problems I would of course
> encourage them to fix the underlying issue (and of course, to enable
> DNSSEC). :) But if they can't/won't/etc ....
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: ma...@isc.org
Joe Baptista
With my business hat on though I can see at least 2 possible use cases for DO=0. The first being related to this thread, "I can't/won't fix/remove the firewall today, I just want my resolver to work." The hapless user in that spot is either going to use another vendor, or go back to the old version of BIND that "works." I know market share isn't a _primary_ concern for BIND, but I would argue that the "go back to old version" answer to this dilemma is something that we should all be concerned about.
I understand - I do anticipate others share your concern.
The other use case that leaps immediately to mind is "We do 42 scintillion DNS queries per second and our bandwidth cost has tripled in the last 3 months! What in the name of J. Jonah Jameson is going on around here?!?"
DNSSEC support is a world wide expense. Not only for the users who deploy it and the registries that support it. But also in bandwidth. If your saying your DNS traffic has tripled thats sounds about right.
Everybody profits and everybody pays.
Since we have Paul's attention here my question is will he incorporate DNScurve into BIND now or does he intend to wait until it becomes an RFC?
regards
joe baptista
Doug Barton
> Doug Barton<do...@dougbarton.us> writes:
>
>> With my business hat on though I can see at least 2 possible use cases for
>> DO=0. The first being related to this thread, "I can't/won't fix/remove the
>> firewall today, I just want my resolver to work."
>
> reasons we fall back to BUFSIZE=512 before falling all the way back to DNS
> (that is, turning EDNS off all together.)
The OP's problem was that the firewall between his resolving name server
and "the cloud" blocks packets with DO=1. Now admittedly this is a
special kind of stupidity on the firewall's part ...
>> In all fairness, I don't have any actual clients telling me that DO=1 is
>> a problem for them, this is pure speculation on my part; ...
>
> yes, i know that, because i'd see the other side of it if it was going on.
Right-O. OTOH discussion/thought about the problem now, before it turns
into a crisis, (probably) can't hurt anything. :)
Doug
Doug Barton
> In message<4C09C562...@dougbarton.us>, Doug Barton writes:
>
> The resolver works. It figures out that it can't make the new style
> queries and falls back to the old style queries. If the user is really
> worried they can turn off EDNS and with that DO.
The OP's problem was that his firewall blocked anything with DO=1.
> It's still a handful of zones that are signed.
But isn't that what we're all working on changing? :)
Mark Andrews
In message <4C0AAD2A...@dougbarton.us>, Doug Barton writes:
> On 06/05/10 07:22, Mark Andrews wrote:
> > In message<4C09C562...@dougbarton.us>, Doug Barton writes:
> >
> > The resolver works. It figures out that it can't make the new style
> > queries and falls back to the old style queries. If the user is really
> > worried they can turn off EDNS and with that DO.
>
> The OP's problem was that his firewall blocked anything with DO=1.
That was the claim. I suspect the reality is something different
and would like to see actual proof that it is not one of the other
firewall issues. This is not to say that there are not firewalls
that choke on DO (when DO was first introduced we saw lookup failures
due to firewalls blocking it) but given named has been sending DO
for years it is strange to get a complaint about DO now.
> > It's still a handful of zones that are signed.
>
> But isn't that what we're all working on changing? :)
>
> Doug
--
Mark Andrews
In message <201006060107....@drugs.dv.isc.org>, Mark Andrews writes:
>
> In message <4C0AAD2A...@dougbarton.us>, Doug Barton writes:
> > On 06/05/10 07:22, Mark Andrews wrote:
> > > In message<4C09C562...@dougbarton.us>, Doug Barton writes:
> > >
> > > The resolver works. It figures out that it can't make the new style
> > > queries and falls back to the old style queries. If the user is really
> > > worried they can turn off EDNS and with that DO.
> >
> > The OP's problem was that his firewall blocked anything with DO=1.
>
> That was the claim. I suspect the reality is something different
> and would like to see actual proof that it is not one of the other
> firewall issues. This is not to say that there are not firewalls
> that choke on DO (when DO was first introduced we saw lookup failures
> due to firewalls blocking it) but given named has been sending DO
> for years it is strange to get a complaint about DO now.
BIND 9.1 sent DO. Every lookup he made would have been slow
(multiple seconds) if DO was a problem for his firewall.
% grep G_DO 9.?.x/lib/dns/*.c
9.1.x/lib/dns/resolver.c: rdatalist->ttl = DNS_MESSAGEEXTFLAG_DO;
9.2.x/lib/dns/message.c: if ((ps->ttl & DNS_MESSAGEEXTFLAG_DO) != 0)
9.2.x/lib/dns/message.c: mbz = ps->ttl & ~DNS_MESSAGEEXTFLAG_DO & 0xffff;
9.2.x/lib/dns/resolver.c: rdatalist->ttl = DNS_MESSAGEEXTFLAG_DO;
9.3.x/lib/dns/message.c: if ((ps->ttl & DNS_MESSAGEEXTFLAG_DO) != 0)
9.3.x/lib/dns/message.c: mbz = ps->ttl & ~DNS_MESSAGEEXTFLAG_DO & 0xffff;
9.3.x/lib/dns/resolver.c: rdatalist->ttl = DNS_MESSAGEEXTFLAG_DO;
9.4.x/lib/dns/message.c: if ((ps->ttl & DNS_MESSAGEEXTFLAG_DO) != 0)
9.4.x/lib/dns/message.c: mbz = ps->ttl & ~DNS_MESSAGEEXTFLAG_DO & 0xffff;
9.4.x/lib/dns/resolver.c: rdatalist->ttl |= DNS_MESSAGEEXTFLAG_DO;
9.5.x/lib/dns/message.c: if ((ps->ttl & DNS_MESSAGEEXTFLAG_DO) != 0)
9.5.x/lib/dns/message.c: mbz &= ~DNS_MESSAGEEXTFLAG_DO; /* Known Flags. */
9.5.x/lib/dns/resolver.c: rdatalist->ttl |= DNS_MESSAGEEXTFLAG_DO;
9.6.x/lib/dns/message.c: if ((ps->ttl & DNS_MESSAGEEXTFLAG_DO) != 0)
9.6.x/lib/dns/message.c: mbz &= ~DNS_MESSAGEEXTFLAG_DO; /* Known Flags. */
9.6.x/lib/dns/resolver.c: rdatalist->ttl |= DNS_MESSAGEEXTFLAG_DO;
9.7.x/lib/dns/message.c: if ((ps->ttl & DNS_MESSAGEEXTFLAG_DO) != 0)
9.7.x/lib/dns/message.c: mbz &= ~DNS_MESSAGEEXTFLAG_DO; /* Known Flags. */
9.7.x/lib/dns/resolver.c: rdatalist->ttl |= DNS_MESSAGEEXTFLAG_DO;
%
> > > It's still a handful of zones that are signed.
> >
> > But isn't that what we're all working on changing? :)
> >
> > Doug
> --
> Mark Andrews, ISC
> 1 Seymour St., Dundas Valley, NSW 2117, Australia
> PHONE: +61 2 9871 4742 INTERNET: ma...@isc.org
> _______________________________________________
> bind-users mailing list
> bind-...@lists.isc.org
> https://lists.isc.org/mailman/listinfo/bind-users
Jan Buchholz
issue this week on the german Linux Tag in Berlin.
What your meaning off firewalls, who looks into packets and block them
if the filter don´t know a flag.
First i´ve fixed the problem with edns no;
Jan
Warren Kumari
On Jun 8, 2010, at 6:26 AM, Jan Buchholz wrote:
> Thanks @all, sorry i was out of office yesterday. I'll discuss the
> issue this week on the german Linux Tag in Berlin.
>
> What your meaning off firewalls, who looks into packets and block them
> if the filter don´t know a flag.
Some "high security" firewalls examine the actual payload of the
packets and validate that the payload follows the spec (at least as
they understand the spec). This sounds like a great win, because it
allows you to make sure that folks aren't tunneling things over other
ports, "protects" your backend from application level attacks (and
attacks on the TCP stack and such) and allows NAT fixups for things
like SIP -- this is often called an ALG (Application layer gateway),
fixups or something similar. Unfortunately they almost always cause
way way more issues than they solve, and cause really really
interesting troubleshooting problems[0]. The firewall has to maintain
a huge amount of state, the ALG is coded for a protocol at a specific
point in time and so doesn't deal with extensions (like edns
apparently :-P), etc.
W
[0]: My favorite instance of this was downloading an ISO of Ubuntu
something or other. I downloaded the ISO and ran 'md5sum' to validate
it -- validation failed so I deleted it and tried again. Validation
fails again. Lather, rinse, repeat.
After a few tries (all over a 1.5mbps DLS line no less) I ended up
copying it over SCP instead of HTTP. Validates fine....
I run 'diff' to see if I can figure out what the hell is going on.
I discover that (in two places in the file) the sequence 0x4772 0x26C7
has mysteriously become 0xC0A8 0x002F. I spend a while poking at
random things (for some reason I had decided it must be bad RAM on the
RAID controller) and end up converting the bytes to decimal -- the
correct one is 71 114 38 199 and then incorrect one is 192 168 0 47...
Wait a minute, that last set of numbers looks *awfully* familiar...
Yup, it's the address of my machine and the other address is the
outside address of the firewall...
Suddenly I realize -- the firewall / NAT device is doing NAT "fixup"
by blindly replacing the "outside" address with the "inside" address
anywhere in the payload... Wheeeeee.....
>
> First i´ve fixed the problem with edns no;
>
> Jan
> _______________________________________________
> bind-users mailing list
> bind-...@lists.isc.org
> https://lists.isc.org/mailman/listinfo/bind-users
--
I had no shoes and wept. Then I met a man who had no feet. So I
said, "Hey man, got any shoes you're not using?"
Mark Andrews
In message <D7C8ADA3-F213-4AE9...@kumari.net>, Warren Kumari wri
tes:
> On Jun 8, 2010, at 6:26 AM, Jan Buchholz wrote:
>
> > Thanks @all, sorry i was out of office yesterday. I'll discuss the
> > issue this week on the german Linux Tag in Berlin.
> >
> > What your meaning off firewalls, who looks into packets and block them
>
> Some "high security" firewalls examine the actual payload of the
> packets and validate that the payload follows the spec (at least as
> they understand the spec). This sounds like a great win, because it
> allows you to make sure that folks aren't tunneling things over other
> ports, "protects" your backend from application level attacks (and
> attacks on the TCP stack and such) and allows NAT fixups for things
> like SIP -- this is often called an ALG (Application layer gateway),
> fixups or something similar. Unfortunately they almost always cause
> way way more issues than they solve, and cause really really
> interesting troubleshooting problems[0]. The firewall has to maintain
> a huge amount of state, the ALG is coded for a protocol at a specific
> point in time and so doesn't deal with extensions (like edns
> apparently :-P), etc.
You wonder about firewall vendors and whether they are doing their jobs
when they don't support parts of the protocol that are a decade old and
are standards track.
EDNS 1999, DO 2001
Mark