Forgot your password?
typodupeerror
The Internet Networking IT

BIND 10 Development Now Fully Underway 76

Posted by timothy
from the ten-is-such-a-nice-round-number dept.
darthcamaro writes "A decade after work first began on version 9 of BIND, the widely deployed open source DNS server, work is now fully underway on its successor, BIND 10. '"One of the goals for BIND 10 is to allow people to customize and extend without too much trouble," Shane Kerr, BIND 10's program manager at the Internet Systems Consortium (ISC), told InternetNews.com.' Sounds good right? Only problem is that it's going to take a bit of time until BIND 10 is actually ready for production — potentially as long as five years!"
This discussion has been archived. No new comments can be posted.

BIND 10 Development Now Fully Underway

Comments Filter:
  • five years! (Score:2, Troll)

    by Jonah Hex (651948)

    potentially as long as five years!

    Why the emphasis on how long it will take? I've had pieces of shit that took longer than that to get ready for "production".

  • What happened to the god of BIND and DNS?
    • Re: (Score:3, Informative)

      by Ethanol (176321)

      He's the president of the company [isc.org] that's doing the work.

      • by Nethead (1563)

        Isn't that vix.com?

        • by wkcole (644783)

          Isn't that vix.com?

          No. vix.com is Paul's personal domain. isc.org is the Internet Systems Consortium, which he heads. ISC is a non-profit that is the custodian for BIND, the reference DHCP implementation, and a few other bits of open source software. It would be at odds with reality to confuse and conflate the two. This is particularly true in regards to actually "doing the work" for developing BIND, since Paul explicitly stayed out of the v9 code, and has publicly referred to the v4 and v8 code as evidence against his progra

      • by oldhack (1037484)
        Oh. Well, that makes me feel ... less old.
      • Re: (Score:3, Insightful)

        by rs79 (71822)

        Paul's doing fine, he and Brian Reid are working together at ISC these days. Brian, if you recall, is the guy who originally funded Paul to take the Berkely B-tree stuff and turn it in to usable software (*) while they were at Digital. They also do some load testing stiff on dns servers for the nsf. You can poke around and find their papers if you look.

        (*) for some definition of "usable". I use djb which annoys them both no end.

  • by Zombie Ryushu (803103) on Thursday June 11, 2009 @06:42PM (#28301727)

    Please, Please Please fix the Bind LDAP SDB Backend to allow LDAP Integrated Zones to Dynamic update. LDAP zones are useless right now because DHCP can't update it!

    • Re: (Score:2, Informative)

      by ritcereal (1399801)
      Read the man page for the dhcp3 configuration for 'on commit', 'on release', and 'on expiry' where you can run code to update your LDAP backend with whatever language you want. There's no reason you can't already do this.
  • by glitch23 (557124) on Thursday June 11, 2009 @07:04PM (#28302019)
    because after BIND 10 is done all the distributors must package it for their specific customers. This includes appliance vendors that utilize BIND. Speaking of appliance vendors, the article mentions that DNSSEC could eventually be enabled by possibly clicking a single button in an interface but that will be dependent on the interfaces put on top of BIND. I guess if BIND 10 has its own interface then that could work well but appliance vendors put their own GUI on top of their implementations of BIND and it may not always be as simple as a single button click.
    • Re: (Score:1, Offtopic)

      I am extremely offended by your signature. We are a nation of religious tolerance, first and foremost. Secondarily we can be said to be a Christian people, but there are significant minorities of other belief systems. One of the most important parts of having a constitutional republic (in general) and the separation of church and state clause (in specific) is to prevent such minorities from the 'tyranny of the majority'. We must not allow the erosion of rights for any segment of our population, else we can

  • Feature parity (Score:5, Insightful)

    by TopSpin (753) * on Thursday June 11, 2009 @07:32PM (#28302255) Journal

    Please try not to leave behind useful features. Yes, misfeatures should be abandoned. Sometimes mere obsolescence can move a feature into the misfeature column. However, merely uncommon or obscure != "mis". It requires a pragmatic grownup to detect the difference.

    The feature set begins with BIND 9. Too many major revisions of fundamental systems fail to achieve feature parity and long after the "new" is production solid the user base remains stratified into the (neglected) old and the (indifferent) new.

    You must know that after the (entirely reasonable) half decade is spent to produce 10 it will take years to migrate the majority of the user base. The justifiably conservative nature of the BIND user base is such that dropped functionality will retard adoption dramatically. Better to provide parity with BIND 9's feature set and remove one excuse to sit on 9 till 2020.

    Put it on the list of goals, near the top; "Feature Parity with BIND 9". Make it clear that the user base can take this for granted; if BIND 9 can do it, BIND 10 can do it.

    I think you'll find if not a lot more support, at least less resistance. I know you will cut the migration period dramatically.

    • by pklinken (773410)
      Congratulations on your 753rd post.
    • Good lord! 2020! We're supposed to have rocket cars by then- why would we need DNS?
      • Re: (Score:3, Funny)

        by fractoid (1076465)
        Do you REALLY want to assign your rocket car's IP address by hand? O.o
        • by yahwotqa (817672)
          This DNS thing... I don't think it does what you think it does.
          • by fractoid (1076465)
            Yeah, I somehow misread and thought we were talking about DHCP, but I didn't get around to posting a correction. :P Consider it corrected to:

            Do you REALLY want to refer to your rocket car by a numerical IP address? O.o

  • by Bondolo (14225) on Thursday June 11, 2009 @07:34PM (#28302277) Homepage
    For a program who's core functionality is name -> number why is the configuration guide heavier than my tombstone? If the future of every Internet standard is to become as complicated as DNS after 35 years then I sincerely believe that the Internet is doomed. 114 RFCs (not counting 20 or so additional obsolete RFCs), WTF? DNS RFCs [dns.net]

    By the way, SMTP and IMAP folks, you're way ahead of the game. Your stuff is already reached the point of sublime unusability past which no fully compliant implementation is possible. Well done!

    • by timeOday (582209)
      It does seem like the destiny of every software package is to become gradually overwhelmed with features. Nothing is worse than upgrading your system and finding the configuration format for apache or exim has changed.
    • Re: (Score:3, Interesting)

      by Just Some Guy (3352)

      For a program who's core functionality is name -> number why is the configuration guide heavier than my tombstone?

      Mainly because it's required to do so very much. Yes, my named.conf is very complicated. I don't know how much simpler you could make split-zone DNS for about 30 zones, including masters, slaves, and some dynamic updates. Oh, and TSIG to authenticate request between each pair of servers. And reverse zones. And IPv6. And recursion (but only for one of the views). I mean, it's sort of like Apache's httpd.conf. Sure, it gets twisty, but what could you leave out and still be able to configure the same f

      • by Cyberax (705495)

        BIND is PITA. It's not modular.

        I've switched my entire infrastructure to djbdns and I'm glad I did it.

        For example, I don't need TSIGs because djbdns uses plain rsync over SSH (which utilizes my PKI) for zone transfers. Dynamic updates are performed using simple shell scripts.

        Everything is pretty simple.

        • For example, I don't need TSIGs because djbdns uses plain rsync over SSH (which utilizes my PKI) for zone transfers. Dynamic updates are performed using simple shell scripts.

          That's the kind of simplicity that makes life much more complex. Our master BIND at work accepts updates for the LAN from Windows desktops. Whenever they get their DHCP lease, they say, "hey, user23.lan.example.com is now at 10.0.0.8". BIND dutifully updates its records and relays that message to the slave BINDs. Contrast with your setup: does djbdns even accept dynamic updates like that? If so, what happens when updates start to come in faster than rsync can copy them to the slaves (I can imagine some

          • by Cyberax (705495)

            "Whenever they get their DHCP lease, they say, "hey, user23.lan.example.com is now at 10.0.0.8". BIND dutifully updates its records and relays that message to the slave BINDs."

            Yes, djbdns can process dynamic DNS updates (via special plug in). Yes, it can then relay records to slaves _or_ you can make multi-master replication (if your master is down).

            "If so, what happens when updates start to come in faster than rsync can copy them to the slaves (I can imagine some pretty large zones with tens of thousands o

            • rsync is incremental protocol. _EXACTLY_ like IXFR used in BIND. So there won't be much difference.

              That's not quite right. IXFR is implemented in BIND as a journal playback (O(1)), but rsync has to examine the entire database for changes to propagate (O(n), where n = number of records in the zone).

              And the worst of it - you can't customize BIND without patching source code.

              How do you customize djbdns without patching source code?

              Personally, I switched to djbdns when I found out that I can't have a hidden DNS slave which can correctly work with views (i.e. I wanted a recursive resolver for my LAN with replicated zone).

              What do you mean exactly? You've already made your choice and that's cool, but I'd be willing to bet that BIND does actually support that feature.

              • by Cyberax (705495)

                "That's not quite right. IXFR is implemented in BIND as a journal playback (O(1)), but rsync has to examine the entire database for changes to propagate (O(n), where n = number of records in the zone)."

                Yes, you are right. But in practice, rsync works fast enough even with zones with hundreds thousands of hosts.

                With djbdns I can easily try to use SVN or git for zone updates.

                "How do you customize djbdns without patching source code?"

                djbdns is incredibly modular. It's a collection of small utilities, each doin

                • Yes, you are right. But in practice, rsync works fast enough even with zones with hundreds thousands of hosts.

                  OK, but it still doesn't address the problem of when to trigger a sync. Do you run it from a cron job, or can you script it from the dynamic DNS update program? Actually, I guess you could relay the update request out to the slaves and not worry about syncing it often. Is that possible?

                  With djbdns I can easily try to use SVN or git for zone updates.

                  That's not a bad idea.

                  No, it doesn't. It works only if I disable views. I checked BIND source to be sure.

                  What I meant was that I didn't understand what you were trying to do and so can't confirm or refute it on my own.

                  • by Cyberax (705495)

                    "OK, but it still doesn't address the problem of when to trigger a sync. Do you run it from a cron job, or can you script it from the dynamic DNS update program? "

                    You can script it (I use a cron job, because I don't need fast updates).

                    "Actually, I guess you could relay the update request out to the slaves and not worry about syncing it often. Is that possible?"

                    Yes, there's actually a plugin which uses SMTP/POP3 to broadcast changes (which is a neat idea, IMO). It's also easy to pipe notifications through SS

                    • by Phroggy (441)

                      Admittedly I've never set up multiple views, but something about that definitely doesn't sound right.

                      Do you mean to say that the recursive resolver on the LAN view is forwarding all queries to your zone's nameservers in the US, or do you mean that it's forwarding queries for your domain to your zone's nameservers while resolving everything else exactly the way you want it to?

                      If the former, you've obviously misconfigured it, because it wouldn't do that if you didn't tell it to.

                      If the latter, that definitely

                    • But it doesn't use it to resolve queries from my LAN! Instead, it forwards requests to my zone's nameservers. Even though it has a full copy of my zone.

                      Phroggy's right: that's definitely possible. I do this at home where I slave my office's zones in the "private" view.

                    • by Cyberax (705495)

                      Yes, it's the latter.

                      BIND tries to resolve names in my zone just as usual (i.e. by asking authoritative nameservers), even though it has a complete copy of my zone.

                      Setting up two nameservers would work, but it's incredibly clumsy.

                      Also, I was able to automate some other tasks with djbdns. Particularly, handling of split-view DNS for another zone.

                    • by Cyberax (705495)

                      Is your office's BIND slaved to master? Or justs hosts a master zone in the private view?

                      The latter works fine, I used it all the time in BIND before I migrated to djbdns.

                    • Home, you mean? I slave the office zone in the private view.
                    • by Cyberax (705495)

                      Have you checked that it really works while you're offline? If it does, then can you send me named.conf, please?

                    • I can't actually verify it from here at work, but really, it's just a slave zone inside the private view.
                    • by Cyberax (705495)

                      Yes. So do I:
                      =====
                      view "internal" {
                      match-clients {127.0.0.0/8;192.168.20.0/24;};
                      recursion yes;

                      zone "somezone.net" {
                      type slave;
                      file "/var/cache/bind/somezone.net";

                    • That's the beatenest thing. I don't have any idea why it wouldn't.
                    • by Cyberax (705495)

                      The problem is here (view.c, row 1176):
                      =========== /*
                      * If the zone is defined in more than one view,
                      * treat it as not found.
                      */
                      zp = (zone1 == NULL) ? &zone1 :
                      result = dns_zt_find(view->zonetable, nam

    • by metamatic (202216)

      If you don't need the complicated functionality of BIND, you shouldn't use it.

      For instance, DNS caching for a home network can be done using something far smaller and simpler, such as pdnsd.

  • So they finally figured out that djb was right; it took them mere 10 years. If he only didn't have such a... personality. Maybe we would've had something sane by now.

    • Re:Modular design? (Score:4, Insightful)

      by e9th (652576) <e9th@tup o d ex.com> on Thursday June 11, 2009 @08:17PM (#28302673)
      You know, it's a shame that djb couldn't play well with others. qmail & djbdns show that he really understood SMTP & DNS. Unfortunately, his dogmatism, odd coding style & disdain for comments, and his weird license (until he PDed them) kept both those products from evolving as they deserved.
      • Re:Modular design? (Score:4, Informative)

        by stinerman (812158) <nathan.stine@NOSpaM.gmail.com> on Thursday June 11, 2009 @11:35PM (#28304081) Homepage

        Weird license?

        IIRC, his code was unlicensed. DJB believed that you didn't need a license to run a binary, compile source, etc.; this is debatable. However, you do need a license to distribute someone's copyrighted works; this is very obvious. Of course, as you say he disclaimed the copyright, so the point is now moot.

        • Re: (Score:2, Interesting)

          by e9th (652576)
          Before he placed it into the public domain, his qmail site had a wonderful "Information for Distributors" page. Maybe not technically a license, but when the copyright holder says

          If you want to distribute modified versions of qmail (including ports, no matter how minor the changes are) you'll have to get my approval. This does not mean approval of your distribution method, your intentions, your e-mail address, your haircut, or any other irrelevant information. It means a detailed review of the exact package

      • Re: (Score:3, Insightful)

        by metamatic (202216)

        What killed my use of djb's stuff wasn't any of those things; it was the dependency on his daemontools replacement for /etc/init.d. (Even today, the djbdns FAQ tries to steer people away from using djbdns without daemontools, and only supplies a half-assed script to install manually.)

        • by e9th (652576)
          daemontools per se is actually a pretty cool package, and an example of what I was trying to say. If Dan had made it easier to repackage, or even simply allowed you to configure alternatives to /command and /service, more people might be using it today.
  • by nnet (20306)
    Please disable ipv6 mapped ipv4 AAAA records. Thank you.
  • Bind 10 was written in Perl 6!

  • ...BIND Vista

    If the schedule slips the could call it BIND Forever.

    Were talkin a name server here, Most major OS rewrites have been done in less time.

  • Too late (Score:3, Interesting)

    by mseeger (40923) on Friday June 12, 2009 @04:19AM (#28305285)

    Hi,

    my personal opinion is, that BIND 9 already lived too long and BIND 10 started much too late. If you have to operate huge installations (>250.000 Zones), BIND 9 is close to unuseable.

    Example: Starting BIND 9 with 350.000 Zones already consumes the complete service window (2 hours) we have for works concerning the hardware. You can't even shave off much time by having all zone files on a ram disk (about 10% less time). BIND 9.6 utilizes a single core for 2 hours just to parse and load the information. For comparison a different (comercial) product imports the (same) complete configuration in about 90s (from disk, BIND 9 format) and takes about 4s for start afterwards. I know there are workarounds for BIND, but they come with high operational costs.

    BIND is (IMHO) mainly a reference implementation. It has to implement everything in one single product and suffers the usual penalties for it. I still use BIND 9 myself for several purposes since it has a some advantages too (mainly, that it is OSS).

    Sincerely yours, Martin

    P.S. If there is any interest, i can post some benchmarks and scripts which i used to run them....

    DISCLAIMER: I'm working for a company that is selling DNS products. So i'm not to be considered a neutral party :-). But since i'm doing this for 15 years now, i consider myself at least an experienced biased party.....

    • Out of curiosity, would that still hold true for database backends? If you multiple zones point at the same file, does it have to parse that file for each zone or can it parse it once and cache it? Would converting them to dynamic zones so that BIND stores the data in its journal format make a difference one way or another?

  • Haven't these guys heard of Agile?

"Consequences, Schmonsequences, as long as I'm rich." -- "Ali Baba Bunny" [1957, Chuck Jones]

Working...