Re: Selection of Appropriate Local SMTP Relay
Randy Bush wrote:
I hear it was one of the first Internet's ideas - to declare well-known addresses in addition to the well known ports.
and notice that, in general, it did not catch on. and in the few places it did, it's been a continual pain in the ass. bad-idea fairy strikes again.
Well, the idea to attach FQDNs to services, not to computers, is pretty sound; and could (if implemented) concievably replace lots of ad-hockery like MX RRs, "virtual" HTTP servers, etc, etc. WKS RRs weren't doing that - the boundary between host names and service names was still in place; so WKS RRs were simply redundant. Following the logic of lazylution they quietly died off. --vadim
WKS RRs weren't doing that - the boundary between host names and service names was still in place; so WKS RRs were simply redundant.
Really, the point is, not enough domains implemented -non-standard- WKS... If you didn't need to -remap- the port, why bother ? And, in case you can answer -that- question: Kerberos locks on WKS didn't actually -work- till -years- later. :}
--vadim
Funny, I've been following this thread for a while, and nobody mentioned -- yet -- 2052 and SRV RRs ... if memory serves, it's written in this RFC that the purpose of SRV was to get rid of / improve some of WKS RRs problems. Anyhow, to put some op content here, Microsoft makes heavy use of SRV RRs in Win2K, hence their customers, and what I mean here is significantly increased load on DNS servers because of lots of port remaping setup between peer servers and services -- ldap, kerb, whatever -- so far so good for the traffic (core / professional) transit -- end-to-end --, but I'm afraid there will be a lot of configurations pounding on innocent DNS (ISP/Telco) servers to either try to update or get updates ... hence the enhanced traffic ... I have few metrics but I guess there will be plenty in less than a quarter. Anybody logged yet significant amounts of _default_first_site_name_kerbereos or _default_first_site_name_ldap failed queries / updates to their nameservers ? - Olivier. Richard Irving wrote :
WKS RRs weren't doing that - the boundary between host names and service names was still in place; so WKS RRs were simply redundant.
Really, the point is, not enough domains implemented -non-standard- WKS...
If you didn't need to -remap- the port, why bother ?
And, in case you can answer -that- question:
Kerberos locks on WKS didn't actually -work- till -years- later.
:}
--vadim
participants (3)
-
Olivier Gerschel -
Richard Irving -
Vadim Antonov