RE: DNS-based RR mechanisms
"Abbie Barbir"<[email protected]> Thu, 11 Apr 2002 12:03:49 -0400
| Newsgroups | gmane.ietf.cdi |
|---|---|
| Message-ID | <[email protected]> |
This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01C1E172.7805FFE8 Content-Type: text/plain; charset="iso-8859-1" Adullah, Load balancing uses request routing techniques. If you really do not like the wording of some sections, I invite you to suggest an alternative before the deadline for the Last Call. Please post the alternative text to the list. Thanks Abbie -----Original Message----- From: Abdullah Rayhan [mailto:[email protected]] Sent: Wednesday, April 10, 2002 2:40 PM To: [email protected] Subject: DNS-based RR mechanisms I find the discussing in the request-routing known mechanisms not very accurate when describing the possible approaches in which RR is implemented to start with. Also the title is supposed to reflect the content but it goes as "known" while the content presents possible cases. The discussion is interesting in itself but does not reference how things are really implemented, For example, sec 2.3 talks about how multiple specialized RR-DNS servers are to be used to distribute complex decisions from single server. The motivation behind this, I believe, is load balancing but this is poorly said and primitive in application. RR-DNS server will perform the same functionality as a DNS server. The difference is being authorative over what domain which is the CN or the content provider if you like. Another difference might be is how load balancing is achieved through more elaborate mechanisms than being presented. Load balancing between surrogates or multiple RR-DNS servers? That needs to be said clearly as well. In the same section it goes on to mix NS and CNAME records without any refs. As a matter of fact, this is not allowed in DNS. You need an A record to go with NS record; DNS experts correct me if I am wrong. The example mentioned does make sense but the discussion should be around load balancing and not RR mechanisms. The discussion goes on in sec 2.3.1. The NS redirection is part of the DNS hierarchy and not RR. What this means is that the CN zone file can be a subdomain of the original provider domain. For example, news.soandso.com, news is resolved by CN RR-DNS server via DNS redirection rather than soandso.com. The same argument applies to section 2.3.2 when requests are redirect via DNS to the CN domain but CNAME are not returned, NS and A records are. The discussion of Anycast is theoretical and I doubt if there is any DNS implementation that makes use of it. It is not even documented and we dont know how it works. It is better if is dropped from the draft. I hope that the authors would take these comments into consideration to clarify the presentation of Section 2. Comments for other sections will follow at a latter time so not to have long emails. my 2 cents. regards Abdallah Rayhan ------_=_NextPart_001_01C1E172.7805FFE8 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"> <HTML> <HEAD> <META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; = charset=3Diso-8859-1"> <META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version = 5.5.2655.35"> <TITLE>RE: DNS-based RR mechanisms</TITLE> </HEAD> <BODY> <BR> <P><FONT SIZE=3D2>Adullah,</FONT> </P> <P><FONT SIZE=3D2>Load balancing uses request routing = techniques.</FONT> </P> <P><FONT SIZE=3D2>If you really do not like the wording of some = sections, I invite you to suggest an alternative before the deadline = for the Last Call.</FONT></P> <P><FONT SIZE=3D2>Please post the alternative text to the list.</FONT> </P> <P><FONT SIZE=3D2>Thanks</FONT> </P> <P><FONT SIZE=3D2>Abbie</FONT> </P> <BR> <BR> <P><FONT SIZE=3D2>-----Original Message-----</FONT> <BR><FONT SIZE=3D2>From: Abdullah Rayhan [<A = HREF=3D"mailto:[email protected]">mailto:[email protected]</A>]</F= ONT> <BR><FONT SIZE=3D2>Sent: Wednesday, April 10, 2002 2:40 PM</FONT> <BR><FONT SIZE=3D2>To: [email protected]</FONT> <BR><FONT SIZE=3D2>Subject: DNS-based RR mechanisms</FONT> </P> <BR> <BR> <P><FONT SIZE=3D2>I find the discussing in the request-routing known = mechanisms</FONT> <BR><FONT SIZE=3D2>not very accurate when describing the possible = approaches in</FONT> <BR><FONT SIZE=3D2>which RR is implemented to start with. Also the = title is</FONT> <BR><FONT SIZE=3D2>supposed to reflect the content but it goes as = "known" while</FONT> <BR><FONT SIZE=3D2>the content presents possible cases. The discussion = is</FONT> <BR><FONT SIZE=3D2>interesting in itself but does not reference how = things are really</FONT> <BR><FONT SIZE=3D2>implemented,</FONT> </P> <P><FONT SIZE=3D2>For example, sec 2.3 talks about how multiple = specialized RR-DNS</FONT> <BR><FONT SIZE=3D2>servers are to be used to distribute complex = decisions from</FONT> <BR><FONT SIZE=3D2>single server. The motivation behind this, I = believe, is</FONT> <BR><FONT SIZE=3D2>load balancing but this is poorly said and primitive = in application.</FONT> <BR><FONT SIZE=3D2>RR-DNS server will perform the same functionality as = a DNS server.</FONT> <BR><FONT SIZE=3D2>The difference is being authorative over what domain = which is</FONT> <BR><FONT SIZE=3D2>the CN or the content provider if you like. Another = difference</FONT> <BR><FONT SIZE=3D2>might be is how load balancing is achieved through = more elaborate</FONT> <BR><FONT SIZE=3D2>mechanisms than being presented. Load balancing = between surrogates</FONT> <BR><FONT SIZE=3D2>or multiple RR-DNS servers? That needs to be said = clearly as well.</FONT> </P> <P><FONT SIZE=3D2>In the same section it goes on to mix NS and CNAME = records</FONT> <BR><FONT SIZE=3D2>without any refs. As a matter of fact, this is not = allowed</FONT> <BR><FONT SIZE=3D2>in DNS. You need an A record to go with NS record; = DNS experts</FONT> <BR><FONT SIZE=3D2>correct me if I am wrong. The example mentioned does = make</FONT> <BR><FONT SIZE=3D2>sense but the discussion should be around load = balancing</FONT> <BR><FONT SIZE=3D2>and not RR mechanisms.</FONT> </P> <P><FONT SIZE=3D2>The discussion goes on in sec 2.3.1. The NS = redirection</FONT> <BR><FONT SIZE=3D2>is part of the DNS hierarchy and not RR. What this = means</FONT> <BR><FONT SIZE=3D2>is that the CN zone file can be a subdomain of the = original</FONT> <BR><FONT SIZE=3D2>provider domain. For example, news.soandso.com, news = is</FONT> <BR><FONT SIZE=3D2>resolved by CN RR-DNS server via DNS redirection = rather</FONT> <BR><FONT SIZE=3D2>than soandso.com. The same argument applies to = section 2.3.2</FONT> <BR><FONT SIZE=3D2>when requests are redirect via DNS to the CN domain = but</FONT> <BR><FONT SIZE=3D2>CNAME are not returned, NS and A records are.</FONT> </P> <P><FONT SIZE=3D2>The discussion of Anycast is theoretical and I doubt = if</FONT> <BR><FONT SIZE=3D2>there is any DNS implementation that makes use of = it.</FONT> <BR><FONT SIZE=3D2>It is not even documented and we dont know how = it</FONT> <BR><FONT SIZE=3D2>works. It is better if is dropped from the = draft.</FONT> </P> <P><FONT SIZE=3D2>I hope that the authors would take these = comments</FONT> <BR><FONT SIZE=3D2>into consideration to clarify the presentation of = Section</FONT> <BR><FONT SIZE=3D2>2. Comments for other sections will follow at a = latter time</FONT> <BR><FONT SIZE=3D2>so not to have long emails.</FONT> </P> <P><FONT SIZE=3D2>my 2 cents.</FONT> </P> <P><FONT SIZE=3D2>regards</FONT> <BR><FONT SIZE=3D2>Abdallah Rayhan</FONT> </P> <BR> <BR> <BR> <BR> </BODY> </HTML> ------_=_NextPart_001_01C1E172.7805FFE8--