mail-ng
[Top] [All Lists]

taxonomies

2004-02-06 00:24:44


On 2/5/2004 12:26 PM, Paul Hoffman / IMC wrote:

In specific, I would love to see the lists broken down into 
mostly-separable parts of the next-generation protocol (such as 
"message transport", "message format", "message display", and so on). 
Doing so will help those who are new to designing systems from 
scratch to see that it isn't all one hunk, even though it may appear 
to be.

I agree that taxonomies are useful discussion tools, to the point that I
think we should spend some time on the subject here.

The simple cut is that mail is about:

 + transport - the global network of disconnected mailboxes and
   processes, and how the mail is moved between them

 + applications - the content that is moved across the network

There are a bunch of sub-areas in each of those major hierarchies:

 - transport
   + node addressing (mailbox format)
   + process addressing (app-specific process at destination mailbox)
   + routing
   + transport authentication
   + transport 'headers' (XML versus [...])
   + protocol syntax (verbs, parameters, options, [...])
   + error-reporting mechanisms
   + extension registries
   + hop-by-hop transfer-extension negotiation mechanics
   + end-to-end application-extension negotiation mechanics
   + [...]

 - applications
   + human-to-human comms (presentation format, encodings, etc.)
   + machine-to-machine application $a (eg, ~syslog writes via email)
   + machine-to-machine application $b (eg, ~chess-by-mail)
   + machine-to-machine application $[...] (whatever)
   + extension $e mechanics (eg, greylist update mechanics)
   + extension $f mechanics (eg, e-postage mechanics)
   + extension $[...] (whatever)
   + [...]

Each of those could be debated and discussed for months by teams of very
smart people, but you really need cross-pollenation to drive any one area
very far. For example, it is absolutely critical to design the common
applications and extensions in order to fully grok what the transport
requirements are going to look like. From that PoV, taxonomy helps folks
to think about the related areas of impact.

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/


<Prev in Thread] Current Thread [Next in Thread>