Re: [idn] New protocol proposal: IDNRA

[Dave Crocker <dhc@dcrocker.net>]

At 10:59 PM 8/27/00 -0400, John C Klensin wrote:
> Deploying a new application is, due to the
> nature of the
> Internet and the end to end principle, fairly easy.  
> Upgrading
> one has proven to be much harder (e.g., MIME and the SMTP
> extensions offered a lot to users, but I'd guess we are still
> well under 90% penentration of competent and conforming
> implementations nearly seven years out).  And changing
> infrastructure --and the DNS and its protocols is certainly
> infrastructure-- is much worse, since both the servers/resolvers
> and the applications need to be upgraded.  


This observation is the key to discussions like these.

• A new application can widely deploy in 6months to 1 year, rather easily.

• Upgrading a massively installed base is a factor of 10 worse.  For lucky upgrades, there is incremental utility for each upgraded application.  MIME was an excellent example. 

• Upgrading an infrastructure is another multiple of 2 to 4.  There is no incremental utility.  Utility occurs only after deployment is wide-spread.  That actually causes a disincentive.  SMTP DSN is perhaps an example.

The expectation that the world will switch to a new DNS in a year or
two does not have any historical precedent.  (Gopher vs. web does
not work well as an example since there were multiple experiments at the
time and gopher was merely one, and gopher was not heavily ingrained in
an infrastructure.)