Re: DNS response record format?
From: Roger Abell [MVP] (mvpNoSpam_at_asu.edu)
Date: 07/17/04
- Next message: Roger Abell [MVP]: "Re: How does RoundRobin flag work?"
- Previous message: GLT: "Re: NT 4.0 PDC/BDC In 2003 Domain"
- In reply to: William Stacey [MVP]: "Re: DNS response record format?"
- Messages sorted by: [ date ] [ thread ]
Date: Fri, 16 Jul 2004 19:58:57 -0700
Use of updates would make sense if it were failover one was after,
but it seems to do the load distribution he mentioned updates
would be pretty unworkable/continuous.
-- Roger Abell Microsoft MVP (Windows Server System: Security) MCDBA, MCSE W2k3+W2k+Nt4 "William Stacey [MVP]" <staceywREMOVE@mvps.org> wrote in message news:ezhzsJ0aEHA.3664@TK2MSFTNGP12.phx.gbl... > Just another idea. Instead of writing server code. You could also just > update the records on existing dns server. For example you can call > nsupdate > in your program to update server RRs. Next query will then see the new > updates. In c# you could call nsupdate like you would call "dir" or other > dos program as a process. Perl also has a DNS module that has update > ability. Using either of these methods, you get as creative as you need > as > to how update order is figured out (i.e. cpu %, hits, machine down, etc.) > Just remember that local client cache and intermediate dns server cache > will > "wash" out most of your hard work unless you use a really low TTL - which > also has a price. > > -- > William Stacey, MVP > > "Peter Steele" <psteele@z-force.com> wrote in message > news:uW4nf3zaEHA.1048@tk2msftngp13.phx.gbl... >> What language am I using? I'd *like* to use C# and a nice .NET class that >> abstracts the whole thing, but it appears there is no such class. >> Fortunately, I've dug up some existing C/C++ code that looks like it will > do >> the trick. It seems the DNS structure isn't exactly "user friendly." The >> main reason I'm writing this is to implement a primitive load >> distribution >> system. In my research I discovered the DNS RobinRobin flag that I made >> another post about. It may very well be that we can use that, but I'll > have >> to see it in action to make sure it does what we want. >> >> Peter >> >> "William Stacey [MVP]" <staceywREMOVE@mvps.org> wrote in message >> news:uToXfEvaEHA.1840@TK2MSFTNGP11.phx.gbl... >> > As Roger said, 1034 and 1035 are the best recourses to get started. > Like >> > most things, putting together a simple request/reply is pretty straight >> > forward, however things get really complicated quick if supporting >> multiple >> > RR types. Then add resolver logic, local lookup logic, cache logic, >> update >> > logic, zones, nodes, rrsets, rr's, dnssec, etc, etc. That said, doing >> reply >> > for one or two records without any recursive logic is pretty easy. I >> would >> > start with RRs. >> > >> > I created an abstract RR class and derived all other RRs from it. The >> base >> > class has ownerName, TTL, Type, Class, and RData string. This is >> > common >> > across all RRs. Your derived classes will be specific to its type. So > an >> A >> > record will have a constructor that also takes an IPAddress, an NS > record >> > will have DName string, an SOA record will have many args, etc. The > first >> > challenge you run into is parsing. You need to parse the byte[] >> > request >> > into sections and parse the request out of it. That is ok. You then > need >> > to see what record type is requested and parse the rdata bytes >> > correctly >> for >> > the type. You also need to figure out name compression logic which is >> > a >> bit >> > tricky. Domain names and domain labels can be scattered around the >> request >> > and the reply ( if your also doing a client resolver.) Servers are >> expected >> > to do name compression for replies also. Your server also needs to > parse >> > text records to build the internal Zone/RR memory structure. Your RR >> > classes also need to be able to parse their data types back into a > byte[] >> to >> > build the reply. This is all pretty easy to hard code for very simple >> need, >> > but to make it general and real-world requires a lot of work. All this >> only >> > scratches the surface of what you will need to do if you take the blue >> pill >> > (or is that the red one?). Good luck with your project. hth. BTW, > what >> > language will you be using? Cheers. >> > >> > -- >> > William Stacey, MVP >> > >> > "Peter Steele" <psteele@z-force.com> wrote in message >> > news:eYiq7vraEHA.1656@TK2MSFTNGP09.phx.gbl... >> > > Can anyone point me to a good source, preferably with programming >> > examples, >> > > describing the format of a DNS response record. I need to write a > simple >> > DNS >> > > server that is used to resolve just a few special names in our >> > > domain. >> > I've >> > > got the basic code infrastructure designed but what I'm having >> > > problem >> > with >> > > is how to create the response record. Any pointers would be > appreciated. >> > > >> > > >> > >> >> >
- Next message: Roger Abell [MVP]: "Re: How does RoundRobin flag work?"
- Previous message: GLT: "Re: NT 4.0 PDC/BDC In 2003 Domain"
- In reply to: William Stacey [MVP]: "Re: DNS response record format?"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
