🚀 go-pugleaf

> Can someone clarify for me how layer 7 can be standardized without
> knowing the application?   I think the difficulty in making higher
> layer protocols is that you can't optimize them for a variety of
> services.  I'm working from a context of metropolitan area networks
> (basically grown from LANs) where one hopes to have integrated
> services (data, voice, video - the usual catch-words).  You can't
> even run these all through the same MAC (physical/data-link) level
> never mind transport and session layers.
> 
> What do these proposed layer 4-7 protocols do, and what kinds of traffic
> do they apply to?
> 
> -Mark Garrett
> Bellcore

There are "classes" of application that are known.  For example, File
Transfer and Access (FTAM), electronic messaging (X.400), etc.

If you look at the function of each layer, you can see that it is
possible to standardize a lot of the common functions that are
supposed to be done at a particular layer.

Session basically manages a transport session and de-multiplexes a
transport connection a little further.  (Transport gets you to a
protocol and session gets you to the application)  There is also the
possibility of checkpointing a session to avoid having to restart at
the beginning in the event of a transport failure.

Presentation is supposed to manage data types between the two
applications.  ASN.1 (part of the presentation protocol) defines an
encoding for data that is machine independent (perhaps in a more
complex way than is necessary).  Character sets, etc. are also handled
at this layer.

Application is supposed to define a set of common primitives for
classes of applications.  FTAM defines the protocol for accessing and
manipulating files.  CASE defines a set of Common Application Service
Elements from which other application layer protocols can be built.
Other application layer protocols in the works are a "virtual
terminal" protocol for remote login and a "directory services"
protocol for accessing a network directory.

It isn't possible to define a standard for every application yet to be
conceived, but it is possible to define them for the common classes of
application.  There will be a lot of vendor specific protocols
providing services which aren't standardized.  Some of these vendor
specific protocols may eventually become the basis for new standards.
With user groups such as the GM-MAP group, standards can gain
acceptance and even convince some vendors to support the standards.

As a disclaimer, I don't necessarily feel that the OSI standards are
better than other network protocols, but it appears that they are
gaining acceptance.  A LOT of vendors have announced support and a
relatively large number have shown implementations at such trade shows
as NCC and Autofact.

Hope this helps,
			Doug McCallum
			{cbosgd, hao, ima}!ico!dougm