Re: SMI clarification

[Olivier Miakinen <[e-mail]>]

Harrie Hazewinkel wrote:
> 
> Hi Olivier,
> 
> Olivier Miakinen wrote:
> >
> > David Perkins wrote:
> > >
> > > Below is a small example MIB. Is this MIB valid or not, you be the judge.
> > > The interesting characteristic about the MIB is that it contains a table,
> > > t2Table, whose index clause is a little unusual. If you think its valid,
> > > then let us know, why so. If you think it is invalid, then what must be
> > > changed to make it valid.
> > >
> > (stuff deleted)
> > >
> > > t2Entry OBJECT-TYPE
> > >     SYNTAX      T2Entry
> > >     MAX-ACCESS  not-accessible
> > >     STATUS      current
> > >     DESCRIPTION
> > >         "Row for second table"
> > >     INDEX       { t1AltIndex, t1AltIndex }
> > >     ::= { t2Table 1 }
> > >
> > (stuff deleted)
> >
> > I'm interested in having the answer to your question, and I think
> > that people who responded didn't see THE point. Then I cut extra-stuff,
> > in order for the `INDEX { t1AltIndex, t1AltIndex }' to appear alone.
> >
> 
> According to the SMI (seen from a syntax point of view) this is a valid definition.
> Only semantically there could be something wrong with. But I see no objections against
> this, unless the original writers had somethings else in mind.
> 
> But what is according to you THE point???

THE point is to have the same attribute, twice in the INDEX clause.

I suppose that both values are different, because it wouldn't be
necessary to specify the attribute twice if it was always the same value.
I mean, you can have an instance of t2Sum that is: t2Sum.7.15, where
7 is one t1AltIndex value, and 15 is another t1AltIndex value.

As an other example, if the INDEX clause was:
   INDEX { t2Sum, t2Sum }
then I would say this would be impossible, for the following reason:
If you get t2RowStatus.7.15, you can't know if corresponding t2Sum
is 7 or 15.

But in that particular case, the attribute t1AltIndex doesn't belong to
t2Entry, then it doesn't really matter: it is approximatively equivalent
to INDEX { Integer32, Integer32 }.


I'm sorry I'm not very fluent in english, and maybe my explainations
aren't very clear. But if I didn't make too many mistakes, you should
have seen that I have MY opinion about this syntax, but I would like
to have the opinion of SNMP gurus. I also think this could be clarified
in RFCs, maybe in forbidding to have the same index twice in an INDEX
clause.


-- 
              \\\\                                ////
              (o o)                              (o o)
+----------oOO--(_)--OOo--------------------oOO--(_)--OOo----------+
| mailto:[e-mail]         Phone: [phone] |
| http://www-ism.bull.com/ism            Fax:   [phone] |
+-------------(_)-(_)--------------------------(_)-(_)-------------+