Re: [apps-discuss] [Uri-review] Provisional URI registrations

["t.petch" <[e-mail]>]

----- Original Message -----
From: "Michael Wojcik" <[e-mail]>
To: <[e-mail]>; <[e-mail]>
Cc: "t.petch" <[e-mail]>
Sent: Wednesday, November 24, 2010 4:30 PM
> From: t.petch [mailto:[e-mail]]
> Sent: Wednesday, 24 November, 2010 03:40
>
> Who needs a definition?  I did a quick 'literature search' and found
> plenty of references to tn3270 as a URI but no definition; the
examples
> were all of the form scheme :// host : port No parameters, queries,
> fragments or anything fancy.  On reflection, this seems right; tn3270
> is more sophisticated than many IETF protocols and if it wants
> something, it negotiates it, so no need for a complex URI to specify
> things in advance.

There are many things in TN3270 that can't simply be negotiated, and
that could usefully be represented in a URI. LU name, terminal type and
model, and printer association, for example; those are all things that
TN3270 clients typically let the user configure.

<tp>
Yes we could, but I would be averse to producing a specification to which
noone actually conforms.  That is, if TN3270 were new and we were embarking
on a URI, we would have a free choice, but instead, we have a decade of history,
and I do not know what that is; in a brief search.I could not find any
parameters
currently and would not want to create a specification that noone has
implemented or will implement.

I see the primary need at present to have an agreed basic definition so the
information stays available in a sensible category in our registries.

Tom Petch
</tp>


The tn3270 URI syntax might also conceivably let the user override the
default capability negotiation, for example to force Classic TN3270
rather than TN3270E, or request a non-default subset of TN3270E
functions. That would be useful for testing, among other things.

And while there's no obvious use for path, fragment, and query URI
components in the tn3270 scheme, I can easily imagine an implementor
using them for TN3270 extensions, such as using the path to select an
initial transaction.

> One could write a universal URI definition for such schemes scheme
host
> port means set up a connection of protocol 'scheme' to 'host', default
> port from IANA, but, arguably, we already have all we need in RFC3986.

I think we still need at least a basic tn3270 definition, if we're going
to consider the scheme well-defined. (Whether its status should be
changed from "provisional" to "historic" in the absence of such a
definition is another question; I have no strong feelings about that.)
There's a lot of variation among TN3270 implementations and the relevant
RFCs don't provide enough direction.

--
Michael Wojcik
Principal Software Systems Developer, Micro Focus



This message has been scanned for viruses by MailController -
www.MailController.altohiway.com