[ipv6-ops] Re: So why is "IPv4 with longer addresses" a problem anyway?

[Ted Mittelstaedt]

On 5/26/2010 2:32 PM, Mark Smith wrote:
> On Wed, 26 May 2010 11:45:32 -0700
> Aaron Hughes<aaronh at bind.com>  wrote:
>
>> Doug,
>>
>> - Survivability
>> - End to end reachability
>> - Guaranteed uniqueness
>> - Some added security (everyone will be forced to address security without NAT in place)
>> - IPv4 space will become very expensive post RIR exhaustion.
>> - IPv6 allows for longer term architecture (allocations and assignment) planning.
>> - Filtering is easier
>> - Regional aggregation is easier
>> - TE (Traffic Engineering) is easier
>> - You will likely never haver to go back to your provider or registry for more IP resources.
>> - Today, IPv6 transit is free ->  $inexpensive
>> - When the day comes (first IPv6 only reachable content/eyeballs come online, you will reach them.
>> - Client / eyeball / content analysis gives more accurate data
>> - From the end user perspective, real IP addresses for all objects. (no more pay for or small subs)
>>
>> While this is a list of perfectly justifiable reasons, the key factor to selling this to your clients is survivability. This will cost money, it will take time, it will cause problems while migrating, but must be done. We don't run IPX/SPX,
>
> IMHO, if the Internet was running IPX (32 bit network number, 48 bit
> node address), we'd still be running it, and wouldn't now have the
> problem of running out of addresses. It had nearly all the features IPv6
> has, and it had them in the early 90s. I think one of the major reasons
> it didn't survive because IPv4 is what the Internet runs.   Having
> personally gone from virtually not learning IPX and working with it
> quite successfully, to then learning IPv4 and all it's quirks like
> classes, subnet masks, prefix lengths etc., it seems to me that being
> connected to the Internet was more important to people than the
> protocol that operated it and it's drawbacks and complexities. I think
> people also, sometimes very slowly, tend towards efficiency. Once IP was
> deployed in people's networks, and there was no chance of getting rid
> of it (because of this Internet thing), if IP could replace IPX it was
> inevitably going to.
>

No, that's not why IPX was replaced.  IPX was replaced because Novell
didn't open-source it and because of that IPX was viewed by the
network community as being tied to Novell's fortunes - then Novell
crashed and burned.

If Novell had retained USL and done what their original gameplan
was, which was to move NCP on to a UNIX server core, and done it
by first releasing IPX into the Net/2 distribution (so that everyone
could run it) then MacOS X wouldn't exist, and MS server and Netware
today would have had the market split.

Novell also badly underestimated the criticalness of having a
solid client.  ipx.com and netx.exe was perfect for DOS systems
and OK for Win31, as long as you knew to use the updated dlls.

But, when Win 95 came out they didn't have a 32-bit client
and the 16 bit clients were kludges under win95.

Ted