> Is it ok to publish this as an Informational RFC?
I would say NO, unless Informational is "whatever enters the RFC
editors queue". See comments.
Source routing by changing format of A records in DNS is a
_bad_ technical idea. Yet another URN solution is another bad idea.
(Messing with DNS in creative ways to serve content "closer" to clients
may yield the desired results, but this is _not_ the way to do it. And
this does not describe how it is done today)
Ingrid
Comments on draft-jl-pcdp-01.txt:
The draft tries to describe a proposed DNS change that will give
information on "closeness" to a client, based on the client IP address.
No references to any DNS specifications. No discussion of consequences
for changes in specification of DNS records. Changing the format of
A-records is not trivial.
No discussion of other solutions to the closeness problem (and why
those solutions mostly fail).
The author discusses this for FTP and HTTP, but there term "TCP" is not
mentioned. And the scheme proposed seems to live in a CIDR-less world.
Why replace routing, and all the neat things routing do? Why limit the
proposal to FTP and HTTP?
Round robin DNS is described as the only deployed answer to "multiple
server for same object". Not even the common CDN-technologies are
mentioned (apart from total silence on multicast and anycast). Not a
single word on caching.
Where does the idea that multiple hop via multiple ISPs is a problem
come from? As long as the content is delivered, there is no problem.
The most favorable reading of the draft is that the author is trying to
implement URNs in his own way. See URN specs for discussion on why
this particular solution was not chosen.
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:30 MST