One of the nasty features of Gateways is that they are generally forced to
lose information, just because some of it cannot be translated.
This lesson was driven home in spades back in the days of X.400 trying to
Translation always has an imperfect because the two languages do not have
matching concepts imbedded in their words.
Consider ho likely it is for the result of translating from English to French,
then French to German, and then back to English.
What would you bet that they match when compared.
I would rather see the gateway verbatim enclose the gateway input into a
"forwarding" envelope such that the exact input can be restored if needed.
At 23:06 -0800 1/29/04, Chuq Von Rospach wrote:
On Jan 29, 2004, at 3:45 PM, Hallam-Baker, Phillip wrote:
One of the things that has consistently failed in the traditional legacy
infrastructure has been upgrading old protocols.
I look to the BITNET model here. The two protocols don't cleanly interoperate,
but work via gateways.
The incentive to upgrade comes from, among other things, gateways increasingly
adding artificial delays into gateway transfers, reducing the immediacy of
delivery, and eventually decomissioning, forcing the issue. You allow for the
transition, but you don't allow it to carry forward forever.
Of course, it's unlikely we'd be able to finish the migration until after IPV6