Re: HTTP URL to support multiple naming services

On Wed, 1 Feb 1995, Paul Rarey wrote:

> On Feb 1,  5:21, Jon P. Knight wrote:
> Although I haven't digested the IPng papers, I believe it is a safe 
> bet that application(s - or application protocols such as hppt, smtp etc.) will 
> not have to change in their use of TCP in order for IPng to be the encapsulating 
> network layer protocol).

Erm, I'm not so sure.  I haven't tracked the development of IP:ng recently
but at least at one point on the big-internet mailing list there was talk
of having to re-write or at least recompile applications to make use of
IP:ng natively.  Don't forget that on UNIX boxes running BSD-derived code,
there isn't a nice, clean, ``please open me a TCP connection to''.  You have to get your hands dirty with sockets using
SOCK_STREAM or SOCK_DGRAM and that means that your code is littered with
structures that have IP addresses in them.  Maybe someone with
their finger on the IP:ng pulse will say yea or nay on this?

I would have thought that most of the on-the-wire protocols will be ok
though.  The only problems I can think of are if there are any protocols
that specifically leave a space for an IPv4 (or IPX or whatever) address
in their on-the-wire representations.

My point was that if you will have to rewrite your code come IP:ng change
over time (ie: tomorrow :-) ), you might as well do the decent thing and
utilise a single namespace and resolution mechanism for your machines.  Why
would you need more than one?  Maybe I'm missing something obvious here...


Jon Knight, Research Student in High Performance Networking and Distributed
Systems in the Department of _Computer_Studies_ at Loughborough University.
* It's not how big your share is, its how much you share that's important *

Received on Wednesday, 1 February 1995 14:19:24 UTC