> -----Original Message-----
> From: Juergen Schoenwaelder [mailto:schoenw@ibr.cs.tu-bs.de]
> Sent: Wednesday, July 10, 2002 05:01
> To: mrw@windriver.com
> Cc: xmlconf@ops.ietf.org
> Subject: Re: Final BOF Agenda?
>
>
> >>>>> Margaret Wasserman writes:
>
> Margaret> I think (but am not sure) that Andy also wants to talk about
> Margaret> how instance information could be represented in XML.
>
> Just to be clear:
>
> - The smidump xsd output format translates a MIB module (SMIv1/SMIv2)
> into an XML schema which defines XML document formats for instance
> information.
Cool.
> - The smidump xml output format translates a MIB module (SMIv1/SMIv2)
> into an XML document which can be used to process MIB module
> definitions with XML tools. (This is an older project and not really
> relevant in the context of xmlconf - but it is important to not
> confuse them.)
>
> A student is also working on an implementation (which I guess will be
> available soon) which implements a gateway that returns instance data
> in XML conforming to the XML schema generated by the smidump xsd
> translation.
How is the desired instance requested from the gateway? XML query?
> The question which remains to be discussed is how useful such an
> automated translation really is. Sure, you get access to MIB data in
> XML format which is easier to process for many people. Bulk transfers
> become easier since you usually assume TCP as a transport. However,
> with an automated translation, you do not really make any steps
> forward to overcome the semantic mismatch between the low-level
> data-centric MIB models and the more task-oriented models operators
> seem to prefer. This also has to do with naming and references which
> are implemented by some kind of pointers in MIB modules, where you
> can't really hide MIB specifics when doing automated translations.
> Perhaps it is more useful to do half-automated translations where MIB
> augmentations are defined for translation purposed to derive XML
> schemas that are closer to the underlying conceptual model.
>
> /js
>
> --
> Juergen Schoenwaelder <http://www.informatik.uni-
> osnabrueck.de/schoenw/>
Thanks, /gww