90 likes | 210 Views
The Origins of the DNS and Its Namespace. Craig Partridge Subbing for Several More Worthy Folks. In The Beginning. Early host database was a text file Operational nightmare! Rate of change in file was growing exponentially NIC staff overwhelmed
E N D
The Origins of the DNS and Its Namespace Craig Partridge Subbing for Several More Worthy Folks
In The Beginning... • Early host database was a text file • Operational nightmare! • Rate of change in file was growing exponentially • NIC staff overwhelmed • Everyone copied it (nightly!) to get current version • Network overwhelmed • Many opportunities for errors • And we experienced many of them • Fate sharing model wrong • Affected parties couldn’t change database • Flat name space: who gets to have Frodo?
The Domain Name System • A distributed database that implemented a hierarchical namespace that could be delegated • EDU server delegates management of harvard.edu to Harvard • Harvard can sub-delegate internally if desired • Fixed most problems • Maintenance now distributed • Network traffic reduced (after we got bugs out) • Fate sharing (you make errors in your own namespace) • Multiple hosts named frodo
The Roll Out Begins • Namedroppers mailing list starts Mar ‘83 • Initial DNS spec released Nov ‘83 • The first namespace RFC came out in Oct ‘84 • About this point, the NIC starts to permit domain names in HOSTS.TXT • First usable version of BIND (2.0) appears Aug ‘85 • At CSNET, I’m asked to get BIND working and integrated with email
Fall of 1985 - The Pace Quickens • BIND means UNIX folks now fully engaged • Namespace issues crop up • Is the namespace right? (E.g. NET) • Any rules about what you can do in delegated domain? • What do the names mean (esp. for Email) • Email model turns out to be broken • Mail Drops and Mail Forwarders and MAILA queries • With help, I figure out how to make it work and write RFC • Jon Postel recognizes a botch • Mockapetris, Partridge, Nedved, Postel and Crowcroft sent off to develop Mail eXchanger RRs.
MX in brief • MX records allowed us to mail to arbitrary DNS names • No longer did the right sign of an @ have to be a host • In fact, it didn’t even have to be on the network! • Could forward transparently through email gateways • Namespace issues and use become more acute • A meeting called for January 1986 • Hosted by Jake Feinler of SRI • SRI-NIC, BITNET, UUCP Project, CSNET, ISI (IANA), UC Berkeley BIND team • During the meeting the Challenger blew up
DNS Issues • Would it be universal namespace? • CSNET, UUCP and Internet all prepared to adopt one naming scheme (BITNET unsure) • What should the name structure be? • What about X.400/X.500 namespace compatibility? • Mark Horton says we ought to at least not rule out • Postel firmly opposed to compatibility • Provider info in names? • Harvard.cs.net, harvard.csnet.bbn.com, harvard.edu? • Ddn-nic.sri.com or nic.ddn.mil? • Who decides who gets to have a domain name? • Network turf issues, and procedural issues
DNS Resolution • Yes to universal namespace • BITNET rep firmly browbeaten • Naming issues • X.400/X.500 “compromise” • Postel allocates .us to himself to preclude US interworking • But agrees to permit other countries to structure as they wish • Names don’t include service provider info • Harvard.edu • DDN NIC gives out domain names and sets policies
What We Didn’t Think About... • Trademarks • Never came up in the meeting • We did discuss briefly how to resolve two requests for the same name • We decided that first-come first-served was simplest • Provided first requester was competent • Slightly more restrictions on country codes • Had to be in ISO list of two-level abbreviations